2016-17 Flawless

• G-001 THE BRIEF TOLD ME TO TREAT tcdb'S PARENT NOTES LINE AS A CLAIM TO TEST RATHER THAN A FACT TO TRANSCRIBE, BECAUSE CAV-045 WAS BOUGHT ON THIS SEASON. I tested it and ON THIS PRODUCT THE LINE IS TRUE, which makes this a second data point on CAV-045 with the opposite outcome to the one that created it. The line reads "Notes: Distributed in one series (#1-100), all base cards serial numbered to 25" beside "Total Cards: 100". Three checks, none of them the prose: (1) Checklist.cfm?PageIndex=2 on the parent renders NO card rows - it falls through to the set-overview page - so there is no numbered block past card 100 for EXTRACTION_SPEC 5.3 to exclude, which is the exact thing impeccable-bkb-2016-17 found hiding behind an identically shaped line (its 101-160 rows were all flagged AU); (2) Gallery.cfm paginates the parent in exactly ten pages of ten tiles with no page 11 and page 10 ending at card 100; (3) the flag column was swept over all 100 rows and the distinct token set is "SN25" and nothing else, with no tokenless row and no row numbered above 100. base_set_size=100 and base_print_run=25 therefore rest on enumeration, not on the Notes line. https://www.tcdb.com/Checklist.cfm/sid/158400/2016-17-Panini-Flawless?PageIndex=2 ; https://www.tcdb.com/Gallery.cfm/sid/158400/2016-17-Panini-Flawless?PageIndex=10
• G-002 NOIR_ADDENDUM CASE (a), STATED PLAINLY BECAUSE THE BRIEF ASKS FOR IT AS A THIRD DATA POINT ON AN OPEN SCHEMA DECISION. This release is the UNIFORM-RUN case: one base card per number, a flat 100-card run at a single /25, and the four versions (Bronze /2, Double Diamond /10, Triple Diamond /5, Platinum 1/1) are ordinary base-set PARALLELS recorded in parallels.csv. There is therefore NO base_card_tiers.csv, no tiling question, and nothing for migrations/009's deferred check_tier_partition() to reject - the unspellable case did NOT arise here. card_tier is EMPTY on all 100 base rows, so CAV-043's exact-string join has nothing to resolve and tier_id being NULL is correct rather than a defect; this release is NOT a new member of CAV-043's second class (unlike flawless-bkb-2018-19, which carries card_tier values with no tier file). The contrast with the two catalogued sibling seasons is total: both are two-tier Base Gems / Base SP Gems runs with base_print_run=varies-by-tier, 2017-18 at 160 with a tier file and 2018-19 at 159 with card 130 unissued and deliberately no tier file. Nothing was inherited from either.
• G-003 TWENTY-FIVE OF THE NINETY-ONE SET-SCOPE TIERS IN THIS RELEASE CARRY A PER-CARD VARIABLE SERIAL RUN AND THE SCHEMA CANNOT HOLD THAT STATE. Each was enumerated over the tier's own rows rather than sampled, each is PROVABLY serial numbered, and none has a tier-level print run, so per EXTRACTION_SPEC 2.3 each is written print_run=UNKNOWN / serial_numbered=UNKNOWN rather than the modal token - which is indistinguishable in the data from 'we found nothing'. This is the same expressiveness gap CAV-041 records for a fifth consecutive season and it cost 25 rows here, the largest single toll recorded so far. It cost no ROWS (unlike status-bkb-2017-18, where two base-scope tiers had to be deleted outright) only because every variable tier here is set-scoped, where UNKNOWN is legal. THE FULL CENSUS, so the fact survives even though the column cannot hold it: Distinguished Patch Signatures Gold: AU, MEM, SN10, SN8 | Distinguished Patch Signatures Gold Proof: AU, MEM, SN2, SN1 | Distinguished Patch Signatures Ruby: AU, MEM, SN15, SN12 | Dual Diamond Memorabilia Emerald: MEM, SN5, SN4 | Dual Diamond Memorabilia Gold: MEM, SN8, SN10, SN7 | Dual Patch Autographs Emerald: AU, MEM, SN5, SN3 | Excellence Signatures Gold Proof: AU, SN1, SN2 | Excellence Signatures Ruby Gold Proof: AU, SN2, SN1 | Flawless Autographs Gold Proof: AU, SN2 / AU, SN1 | Greats Gold Proof: AU, SN1, SN2, SN3 | Greats Ruby Gold Proof: AU, SN1, SN2 | Horizontal Patch Autographs Gold: AU, MEM, SN3, SN5, SN7, SN9, SN10 | Momentous Patch Autographs Emerald: AU, MEM, SN5, SN3 | Momentous Patch Autographs Gold: AU, MEM, SN10, SN8, SN7, SN5 | Momentous Patch Autographs Gold Proof: AU, MEM, SN1, SN2 | Momentous Patch Autographs Ruby: AU, MEM, SN15, SN11, SN10 | Patches Gold: MEM, SN10, SN8, SN7, SN6, SN5 | Premium Ink Gold Proof: AU, SN1, SN2 | Premium Ink Ruby Gold Proof: AU, SN1, SN2, SN4 | Rookie Autographs Gold Proof: AU, SN1, SN2 | Star Swatch Signatures Emerald: AU, MEM, SN5, SN3 | Star Swatch Signatures Gold: AU, MEM, SN10, SN8, SN7 | USA Basketball Signatures Gold Silver Gold Proof: AU, SN1, SN2 | Vertical Patch Autographs Emerald: AU, MEM, SN3, SN4, SN5 | Vertical Patch Autographs Gold: AU, MEM, SN10, SN8, SN5. A print_run_basis column, or permitting serial_numbered=yes with print_run=UNKNOWN, would convert this spend into recorded fact.
• G-004 A CONTRADICTION OF A GENERALISATION THE SEASON'S OWN SCOPE MEMO STATES AS HOLDING WITHOUT EXCEPTION, AND IT IS NOT A ONE-OFF. Memo 7.5 measured 2016-17 Prestige at '4 of 4 autograph BASE tiers variable, 0 of 2 colour tiers OF THOSE SAME SETS variable' and wrote that this 'confirms the reliable generalisation exactly ... now holding without exception on every product ever measured'. On 2016-17 Flawless the COLOUR tiers of autograph sets vary too. The cleanest single instance, confirmed on a second differently-varied endpoint because it is load-bearing: Distinguished Patch Signatures Gold (sid 162470) renders all 18 of its 18 rows on BOTH ViewSet.cfm and Checklist.cfm, and Checklist.cfm gives "AU, MEM, SN10" on seventeen rows and "AU, MEM, SN8" on DP-RM (Reggie Miller). A three-row sample would have written a clean sourced-looking flat 10. THIRTEEN FURTHER COLOUR tiers vary the same way, making FOURTEEN OF THE RELEASE'S SIXTY non-proof colour tiers variable - Distinguished Patch Signatures Ruby, Dual Diamond Memorabilia Emerald and Gold, Dual Patch Autographs Emerald, Horizontal Patch Autographs Gold, Momentous Patch Autographs Emerald, Gold and Ruby, Patches Gold, Star Swatch Signatures Emerald and Gold, and Vertical Patch Autographs Emerald and Gold - so the exception is structural on this line, not noise. The second half of the generalisation ('nothing at the 1/1 end varies') ALSO fails here: ELEVEN Gold Proof / Ruby Gold Proof / Silver Gold Proof tiers of thirty-one carry mixed SN1/SN2/SN3/SN4, and 14 + 11 = 25 accounts for every variable tier in the release. The memo's measurement of Prestige is not disputed; its extension to 'every product ever measured' is what this release refutes, and 7.5 itself warns that nothing licenses carrying Prestige's rate onto a premium line. https://www.tcdb.com/Checklist.cfm/sid/162470/2016-17-Panini-Flawless-Distinguished-Patch-Signatures-Gold
• G-005 CAV-042 REPRODUCED LIVE, ON ONE URL, IN THE DANGEROUS DIRECTION. Gallery.cfm/sid/162500/...?PageIndex=8 rendered an EMPTY flag column on all ten of its tiles (FA-RR through FA-VD) under the default sort. The SAME url with &sort=0 appended rendered "AU, SN10" on all ten. Writing unnumbered/no off the first reading would have put a confident, sourced-looking 'unnumbered' on ten provably /10 rows. Recorded because CAV-042's measurements so far are Checklist-versus-ViewSet; this is the same defect between two PARAMETERISATIONS OF ONE GALLERY URL, which the caveat does not currently name. https://www.tcdb.com/Gallery.cfm/sid/162500/2016-17-Panini-Flawless-Flawless-Autographs-Gold?PageIndex=8
• G-006 TRUNCATION AS MEASURED HERE RATHER THAN PLANNED AGAINST. (1) CAV-046's count-versus-span tell fired once: Checklist.cfm?PageIndex=1 on Bronze (sid 162458) returned 50 rows spanning card numbers 1-73, so interior rows were dropped and the page was not used for coverage. (2) The parent checklist behaved well: 78 rows spanning 1-78, count EQUAL to span. (3) ?sort=0 IS NOT A DEPENDABLE REMEDY and made things worse here exactly as CAV-046 warns - on the parent it returned 77 rows against the default sort's 78. It did work as a cache-bust on the Gallery flag-column loss in G-005. (4) ViewSet.cfm on SUBSET sids was the workhorse and its ceiling is unpredictable: it returned 51, 77, 81 and 83 contiguous rows on the four base parallels and full renders on most small tiers, while truncating the 82-83 card Flawless Autographs and Premium Ink tiers at various points. Twenty-three Gallery.cfm top-up pages were needed in all. (5) Several ViewSet responses asserted a row count equal to the stated Total Cards while their own last rendered code proved otherwise (Flawless Autographs Gold claimed 83 rows and ended at FA-PNY); the last-code clause was taken and the tally discarded, per EXTRACTION_SPEC.
• G-007 THE UNIFORM SETS HAVE THEIR OWN PRINT RUNS AND sets.csv HAS NOWHERE TO PUT THEM. EXTRACTION_SPEC 2.2 forbids putting the figure in card_count and there is no print_run column, so it survives only in prose here and in each set's notes. Measured over each set's own rows: Excellence Signatures, Flawless Autographs, Greats, Premium Ink and Rookie Autographs all /25 (AU, SN25); Champions, Most Valuable and Top of the Class /5; Triple Double /3; Championship Tag Autographs and Championship Tags /2; Logoman and Logoman Autographs 1/1; USA Basketball Signatures Gold /10. These are properties of the sets, not of parallels, so they are NOT written as parallels.csv rows.
• G-008 THREE SET COUNTS WHERE tcdb AND cardboardconnection DISAGREE AND THE CAV-044 UPWARD-REVISION RULE DID NOT APPLY, so tcdb's base-sid figure was written and the disagreement is recorded instead of smoothed. Dual Patch Autographs: tcdb states 56 and renders 56 DPA- codes, cardboardconnection states 59 - and tcdb's own Gold tier renders 59 while its Emerald renders 64 and its Platinum 65, so 59 is not the largest tier count and the evidence does not single it out. Flawless Autographs: tcdb states 83, cardboardconnection 82; tcdb's tiers render 82 and 83 on different colours and its Emerald's own Gallery pages total 83 TILES against that page's stated 82. Premium Ink: tcdb states 83, cardboardconnection 82, tiers mostly 83. In all three the plausible reading is a variation or short-print layer counted by one source and not the other, and nobody has opened enough of the rosters to say which. https://www.cardboardconnection.com/2016-17-panini-flawless-basketball-cards
• G-009 FIVE SET COUNTS WERE NOT TAKEN FROM tcdb'S BASE SID, and the rule is stated here so it can be audited or reversed as a unit. Rule applied: where cardboardconnection's count exceeds tcdb's base-sid Total Cards AND equals the LARGEST count tcdb itself renders for any tier of that set, the base sid is treated as rows short per CAV-044 and the larger figure is written on two independent legs. Dual Diamond Memorabilia 18 -> 20 (cc 20, Logoman tier sid 162479 renders 20). Horizontal Patch Autographs 32 -> 40 (cc 40, Emerald sid 162516 and Platinum sid 162518 each render 40 over the same H-AD..H-ZLV span). Patches 79 -> 80 (cc 80, Platinum sid 162535 renders 80). Star Swatch Signatures 22 -> 27 (cc 27, Platinum sid 162562 renders 27). Triple Double 14 -> 15 - the strongest of the five and the only one where a specific card is at stake: the base sid renders the numbers 1-4 and 6-15 with no 5, while Triple Double Platinum (sid 162566) states "Total Cards: 15" and renders 1-15 INCLUDING number 5, which is CAV-044's positive test done exactly as it asks. Writing 14 would have recorded a card as unissued that a parallel page shows exists. https://www.cardboardconnection.com/2016-17-panini-flawless-basketball-cards
• G-010 FORTY-FIVE OF THE 164 SETS tcdb FILES UNDER THIS 2016-17 PARENT ARE 2015-16 PANINI FLAWLESS ROWS. They are FILED in sets.csv with tcdb's name reproduced verbatim including its season prefix and a flag in every note, per EXTRACTION_SPEC 5.5, which says list them and records an orchestrator brief being reversed TWICE for saying gap them instead. What is unresolved: card_count is EMPTY on 44 of the 45 - their own sid pages were not opened and cardboardconnection carries only one combined figure for the whole block ("2015-16 Flawless Extras Set Checklist: 146 cards"), not per-row counts - and set_type on all 45 is derived FROM THE SET NAME ('Autographs', 'Signatures', 'Ink' -> auto, otherwise insert), which is a required NOT NULL enum with no sourced value available. Their sids are 162462 plus 603030-603069 and 667267-667270, from the parent's own "Included Sets:" enumeration. Their codes must never be mixed with this release's.
• G-011 THE TWO SOURCES GIVE THE PRIOR-SEASON BLOCK DIFFERENT NAMES AND THE SAME SIZE. tcdb (sid 162462) heads it "2016-17 Panini Flawless - 2015-16 Panini Flawless Update" with "Total Cards: 146"; cardboardconnection heads it "2015-16 Flawless Extras Set Checklist" with 146 cards. 146 is therefore corroborated by two independent sources, but it is a STATED HEADER plus an editorial figure, NOT an enumeration - the fetch layer rendered only a handful of that page's rows (numbers 125, 150, 160 and 167-170 were visible). tcdb's string is what sets.csv carries. https://www.cardboardconnection.com/2016-17-panini-flawless-basketball-cards
• G-012 NEITHER SOURCE PRINTS A BARE 'USA BASKETBALL SIGNATURES' AND NO BARE NAME WAS INVENTED. tcdb catalogues no uncoloured sid for this set - its lowest is 162567 "USA Basketball Signatures Gold" - and cardboardconnection's heading is likewise "USA Basketball Signatures Gold Set Checklist: 10 cards". The sets.csv row is therefore named with the string BOTH sources print, and its six sibling sids (Gold Gold Proof, Platinum, Ruby, Ruby Gold Proof, Silver, Silver Gold Proof) are set-scope parallels of it. THE VISIBLE ODDITY THAT FOLLOWS: a Silver tier at /20 sits under a Gold set at /10, i.e. the nominal parallel is LESS scarce than its nominal base, which is strong evidence that the real structure is a 'USA Basketball Signatures' set whose Gold is one tier among seven. EXTRACTION_SPEC's rule to prefer a source's longer name over inventing a disambiguator, and the CHASE spec's warning that 'the base set is the lowest sid in its group' is an unsafe inference, point in opposite directions here. Left as the sources print it and flagged for an owner decision rather than resolved silently.
• G-013 THE COPY CHECK FIRED ON THIS TIER AND WAS INVESTIGATED RATHER THAN ACTED ON. Sid 162490 states "Total Cards: 65" and renders a roster of 65 EX- codes byte-identical to the base Excellence Signatures set's own 65, carrying "AU, SN5" - the same size AND the same token as the plain Emerald tier (sid 162489). Confirmed identical on ViewSet.cfm and Checklist.cfm, so it is the page's behaviour and not a fetch artifact. Every OTHER Gold Proof tier in this release is a small subset at or near 1/1 (Distinguished Patch Signatures Gold Gold Proof renders 1 card, its Gold Proof 2, Greats Gold Proof 8, Momentous Gold Gold Proof 1), which is why the shape looks wrong. Against that: Flawless Autographs Emerald Gold Proof (sid 162499) is ALSO full-roster at 83 and Premium Ink Emerald Gold Proof at 41, so a large Emerald Gold Proof is not unique and the tier is not obviously a duplicate. print_run=5 is written because that is what the page headed with this tier's name renders on every row; the suspicion is recorded here rather than resolved, because CLAUDE.md's rule is investigate, do not delete. https://www.tcdb.com/Checklist.cfm/sid/162490/2016-17-Panini-Flawless-Excellence-Signatures-Emerald-Gold-Proof
• G-014 Inserts.cfm ON THIS PARENT TRUNCATES BEFORE ITS OWN 'Parallel Sets' ROWS, and the replacement endpoint was better than the original. Inserts.cfm/sid/158400 rendered the heading "Parallel Sets (4)" and then cut before any of the four rows; a second differently-parameterised fetch (?x=1) cut at the same place, so this is the deterministic large-parent behaviour EXTRACTION_SPEC 3.2 records on the Optic parent and the 2016-17 memo 7.1 records on National Treasures, not stochastic truncation. CONSEQUENCE FOR METHOD, worth reusing: the parent's own set-overview page - reached at Checklist.cfm?PageIndex=3, i.e. any out-of-range page index - renders an "Included Sets:" list that returned ALL 164 entries with their sids in ONE fetch, uncut. The arithmetic then closes exactly: 4 base parallels + 1 prior-season Update + 115 current-season set and tier sids = the contiguous block 162458-162577 (120 sids), plus 44 later-added 2015-16 Update I sids at 603030-603069 and 667267-667270 = 164, my own count. BOTH BLOCK BOUNDARIES WERE PROVED rather than assumed: 162457 is "2017 Panini Flawless Collegiate - Rookie Patch Autographs Sapphire" ("Total Cards: 66"), a 2017 FOOTBALL product - a genuine cross-sport adjacency trap on this exact line name, the same shape as the 2018 Flawless USA Signatures Sapphire hazard recorded for 2017-18 - and 162578 is "1994-95 Stadium Club - Super Teams Exchange: Conference Winner New Jersey Devils" ("Total Cards: 7"), hockey. No sid was interpolated and none in the block is unaccounted for. NOT CLAIMED: that tcdb holds everything. https://www.tcdb.com/Checklist.cfm/sid/158400/2016-17-Panini-Flawless?PageIndex=3
• G-015 THE SAME EVIDENCE SHAPE IS CARRIED UNDER TWO DIFFERENT BASIS VALUES ON ONE LINE, AND NOTHING ON EITHER ROW TELLS A READER WHICH. This release writes Bronze channel=FOTL with channel_basis=sourced-presence, off cardboardconnection's "PARALLEL CARDS: Double Diamond #/10, Triple Diamond #/5, Bronze #/2 (1st Off Line), Platinum 1/1." - a parenthetical that PLACES the tier in First Off The Line boxes without using the word exclusive, which is what sourced-presence means. flawless-bkb-2017-18 writes sourced-exclusive for its own Bronze off an equivalent quote, and flawless-bkb-2018-19 writes from-name for a tier actually NAMED 'Bronze FOTL'. Three seasons of one line, three bases, one evidence shape. This build followed evidence-or-omit and took the weaker value; the divergence is named here rather than resolved by promoting or demoting anyone's row. https://www.cardboardconnection.com/2016-17-panini-flawless-basketball-cards
• G-016 cardboardconnection PRINTS A 'PARALLELS CHECKLIST' CARD COUNT PER SET AND IT IS NOT A COUNT OF ANYTHING I CAN RECONCILE, so it was used for nothing. Example: it gives "Distinguished Patch Signatures Parallels Checklist: 22 cards" plus "Distinguished Patch Signatures Gold Proof Set Checklist: 4 cards" = 26, while tcdb's eight tiers of that set render 21+2+18+1+2+21+19+2 = 86. Same for Dual Patch Autographs (cc 66 against tcdb's 64+59+65=188), Dual Patches (65 against 60+59+65=184), Momentous (50+5 against 46+2+44+1+2+49+35+2=181), Signature Prime Materials (64 against 61+56+64+51=232) and Vertical Patch Autographs (39 against 37+32+38=107). This is the EXTRACTION_SPEC 5.4 unit-2 hazard (parallel-tier LINES rather than distinct cards) without a decomposition that works, so no per-tier figure in parallels.csv rests on it and the tier-count cross-check those numbers looked like they would provide was NOT available. https://www.cardboardconnection.com/2016-17-panini-flawless-basketball-cards
• G-017 SEVERAL SETS IN THIS RELEASE RENDER MIXED CODED/ORDINAL NUMBER COLUMNS, WHICH IS THE CHASE PASS'S PROBLEM AND IS FLAGGED HERE SO IT IS NOT REDISCOVERED. Championship Tags (sid 162466) renders 36 bare ordinals and four real codes (CT-PG at display position 14, CT-AH at 19, CT-KT at 37, CT-JW at 39); Logoman (162519) renders 34 ordinals and one LM-JM; Logoman Autographs (162520) renders the numbers 1 and 3-23 with no 2; Dual Patches (162484) renders one row whose number column is the bare string 25 beside 43 DPT- codes; Vertical Patch Autographs Emerald (162575) and Platinum (162577) both begin with a bare "23"; Rookie Autographs Gold Proof (162550) renders bare ordinals 1-15 where its siblings render RA- codes; Flawless Autographs Platinum (162503) begins with a bare "13". Under the CHASE spec's mixed-code rule one real code settles the scheme, so the bare ordinals are display indices rather than printed numbers - but NONE of these was resolved here and no card-level rows were written. insert_cards.csv is deliberately absent from this directory.
• G-018 THE STUDIO LESSON APPLIED: A FLOOR NEEDS A POSITIVE STATEMENT AND THIS LINE HAS THE OPPOSITE. cardboardconnection's product page says "Although the same basic 10-card box format is continued from the 2015-16 set...", which positively asserts a 2015-16 Flawless exists, and the 2016-17 scope memo 3 places the line at tcdb sids 133638 (2015-16) and 115433 (2014-15). Building 2016-17 therefore does not reach this line's floor and none is claimed. Depth below 2014-15 was NOT measured. https://www.cardboardconnection.com/2016-17-panini-flawless-basketball-cards
• G-019 EVERY NON-tcdb FACT IN THIS RELEASE COMES FROM ONE EDITORIAL LINEAGE. cardboardconnection is the only aggregator consulted; checklistcenter, checklistinsider, sportscardchecklist and waxstat were not opened, so no independence test was run for this product as CAV-037 (as corrected) requires per product and per field. beckett.com was not budgeted for, per CAV-033. Marketplace corroboration was not attempted at all - the 2016-17 scope memo 9 notes it is untested for this season, and 2016-17 is the Simmons/Ingram/Brown/Hield rookie year so the raw material exists. blog.paniniamerica.net was not swept for this product. PRACTICAL CONSEQUENCE: the four base-set parallels are source_confidence=multi on the strength of tcdb's row-level flag column plus cardboardconnection; all 91 set-scope tiers are source_confidence=single, on tcdb alone.
• G-020 parallel_family WAS DERIVED, NOT JUDGED, AND NOTHING WAS ADDED TO pipeline/fix_schema.py (CAV-039). The maps were lifted out of fix_schema.py by AST so the repo keeps one definition, and family_of() was applied unchanged. Result: Emerald 22 rows, Platinum 20, Diamond 2 (Double Diamond and Triple Diamond), Bronze 1, and 'other' 50 - the 'other' rows being every Gold, Gold Proof, Gold Gold Proof, Ruby, Ruby Gold Proof, Silver, Silver Gold Proof and Logoman. NO NEW FAMILY WAS MINTED and none is requested: Gold, Silver and Ruby are colour names the ordered FAMILIES list deliberately does not carry, and minting them in one release would split them from the rest of the corpus, which is the CAV-039 trap. Two derivations worth an owner's eye rather than an edit: family_of('Emerald Gold Proof') returns Emerald, which treats a proof treatment as a colour family; and family_of('Double Diamond') and family_of('Triple Diamond') both return Diamond, which is probably right but collapses two tiers with different print runs (/10 and /5) into one family. Checked for the CAV-019 ordered-substring hazard: no name in this release is swallowed by a shorter pattern.
• G-021 THE CHANNEL WAS DECLINED ON EVERY SET-SCOPE TIER, DELIBERATELY. cardboardconnection states a product-level configuration ("Product Configuration: 10 cards per pack, 1 pack per box, 2 boxes per case") and no source consulted attributes a trade channel, box SKU or region to ANY individual tier of this release. Per EXTRACTION_SPEC 2.3 a declined channel is channel=UNKNOWN with channel_basis EMPTY - not 'unsupported', which describes a KNOWN channel you could not source and which migrations/003's basis_iff_channel_known constraint rejects alongside UNKNOWN. channel_confidence is likewise left EMPTY rather than set to a value, per CAV-038, which records 81 corpus rows carrying an illegal value on that unpoliced column. The one exception is Bronze, where cardboardconnection's ladder annotates the tier "(1st Off Line)": channel=FOTL, channel_basis=sourced-presence, channel_confidence=single. https://www.cardboardconnection.com/2016-17-panini-flawless-basketball-cards
• G-022 rookie_flag IS UNKNOWN ON ALL 100 BASE CARDS AND THAT IS UNRESOLVED, NOT 'no'. Neither endpoint renders an RC designation anywhere in this set: the flag column was asked for BY EXAMPLE including RC on both Checklist.cfm and every Gallery.cfm page, and the distinct token set over all 100 rows is "SN25" and nothing else - including on rows that are unambiguously 2016-17 rookies (6 Ben Simmons, 9 Brandon Ingram, 11 Buddy Hield, 32 Dejounte Murray, 45 Jakob Poeltl, 46 Jamal Murray, 48 Jaylen Brown, 87 Skal Labissiere, 91 Thon Maker, 95 Tyler Ulis, 98 Yogi Ferrell). Per CAV-003 most Panini checklist sources do not print RC designations at all, so this is an unpopulated column rather than a set without rookies, and UNKNOWN is the correct value. cardboardconnection prints a roster but no RC marks that were readable here.
• G-023 NO CLAIM IS MADE THAT THIS RELEASE HAS NO VARIATION SET. Errors.cfm was not treated as a discriminator at all, per CAV-020, which records 43 corpus rows marked for exactly that mistake - its silence is worthless and it returns the same sentence on a set that provably has errata. What DOES survive independently: base_set_size=100, the base run being contiguous 1-100 with no repeated number, and no numbered block past 100 - all three rest on the enumeration in G-001, not on any errata endpoint. A b-suffixed variation layer would not surface in a base-run enumeration and none was looked for, so it is unsupported rather than disproved. Note also that base card_number is an INTEGER column, so had a b-suffixed base row existed it would have been a finding rather than a value.
• G-024 THE EXTRACTION_SPEC 5.5 MIRROR WAS NOT RUN. This parent carries 45 rows of PRIOR-season 2015-16 Flawless material (G-011), so the mirror - this season's Update material filed under the 2017-18 parent, where an agent reading only its own parent sid will never see it - is very likely to exist. flawless-bkb-2017-18's own directory was read for questions only and its sets.csv was not searched for 2016-17 rows, and the 2017-18 parent (sid 182810) was not opened by this build. Nothing in this release depends on the answer, because the base run has no holes for an Update block to explain, but the set inventory may be incomplete in that one specific, checkable direction.
• G-025 DELIBERATELY NOT CHASED AND DELIBERATELY NOT FILED HERE. The task brief records that two 2015-16 Flawless printing-plate groups appear under the 2016-17 National Treasures parent and states that this is an EXTRACTION_SPEC 5.5 matter for the nattreasures-bkb-2016-17 slice, which is being built concurrently. This build did not open the National Treasures parent, did not enumerate those groups, and wrote no row for them, so the absence of any 'Printing Plates' set in this directory is a scope boundary rather than a coverage hole. Recorded only so a later reader does not read the absence as an oversight. Note for whoever reconciles the two slices: this parent's own 45 prior-season rows do NOT include any printing-plate group.
• G-026 NOTHING WAS INHERITED FROM EITHER SIBLING SEASON AND THE STRUCTURE REVERSES AGAINST BOTH. 2016-17 is a flat 100-card base run at one /25 with base_print_run=25; flawless-bkb-2017-18 is 160 cards in two tiers (Base Gems 1-100 /20, Base SP Gems 101-160 /5) with base_print_run=varies-by-tier and a tier file; flawless-bkb-2018-19 is the same two-tier shape at 159 with card 130 unissued, varies-by-tier and deliberately NO tier file. Consequences worth stating: (1) this release has no block-scope parallel rows at all, where both siblings scope Bronze/Double/Triple to a tier, so the block-name-collision hazard of EXTRACTION_SPEC 2.3 cannot arise here; (2) the base ladder's NAMES differ too - tcdb prints 'Double Diamond' and 'Triple Diamond' here where 2017-18 prints 'Double' and 'Triple', and cardboardconnection agrees with tcdb in BOTH seasons, so this is the product changing its ladder names rather than a source divergence; (3) the four names that 2017-18 carries as token blocks inside its 101-160 tier (Triple Double, Top of the Class, Most Valuable, Champions) are separate sets with their own sids here, outside the base run. (4) Since base_print_run is an integer and not varies-by-tier, this release does NOT trip the snapshot.py invariant that flawless-bkb-2018-19 still fails on; snapshot.py was not run by this build, per the task brief.

If you know a source that answers one of these, tell us — name the question's label and where to look.