2017-18 Donruss

• G-002 base_print_run is written unnumbered with basis=assumed, NOT sourced. No source states in words that the base cards carry no serial number. The evidence is a within-parent positive discriminative test: tcdb renders no serial token on any base row across eight separate fetches of sid 153326 while rendering SN299/SN199/SN99/SN75/SN49/SN25/SN15/SN10/SN1 on the rows of twelve of this release's fifteen base parallel sids. Per EXTRACTION_SPEC section 2.1 that is strong indirect evidence and therefore assumed, not sourced.
• G-003 Card counts for 22 of the 31 insert and autograph sets are single-sourced to tcdb's "Total Cards" header. cardboardconnection independently states counts for only nine of them (All Clear for Takeoff 15, All-Stars 30, Court Kings 40, Hall Kings 30, Rookie Kings 30, Retro Series 25, Swishful Thinking 10, The Champ is Here 15, The Rookies 5) and tcdb agreed with it on all nine. No second source was reachable for the remaining 22.
• G-004 checklistcenter.com was searched for a 2017-18 Donruss basketball page and none surfaced in the results (which returned its 2018-19 Donruss and 2020-21 Clearly Donruss pages instead). Per reference/caveats.csv CAV-037 no URL was constructed and this is NOT recorded as an absence claim: the page may exist under a slug the search did not surface. The count corroboration checklistcenter usually provides is therefore missing from this release.
• G-005 Four set counts rest on a tcdb header that the fetch could not confirm against a full enumeration, because ViewSet.cfm truncated: Rookie Jerseys (header 100, ~67 rows rendered), Rookie Jerseys Prime (header 100, ~50 rendered), Significant Signatures (header 98, ~60 rendered), and the parallel Significant Signatures Green (header 100, ~80 rendered) and Black (header 99, ~66 rendered). Per EXTRACTION_SPEC these headers have been measured counting five different units, so each is a question rather than an answer until enumerated.
• G-006 SIX SETS CARRY A PER-CARD VARIABLE PRINT RUN THAT NEITHER sets.csv NOR parallels.csv CAN EXPRESS (EXTRACTION_SPEC section 2.3 schema gap: serial_numbered=yes with print_run=UNKNOWN is not representable). Distinct token sets enumerated on tcdb: Dominators Signatures = SN40 x38, SN25 x1 (Bill Russell), SN30 x1 (Chris Bosh); Hall Dominators Signatures = SN99 x23, SN75 x4, SN65 x1, SN50 x1, SN40 x1; Rookie Materials Signatures = SN75 x32, SN150 x1 (T.J. Leaf); Timeless Treasures Materials Signatures = SN30 x13, SN49 x3, SN40 x2, SN50 x2, SN25 x1, SN23 x1, SN99 x1; Timeless Treasures Materials Signatures Prime = SN10 x13, SN5 x8; Jersey Series Prime = SN10 x46, SN4 x1. The figures are carried verbatim in each set's notes. Writing the modal token as a flat print run would have been false on every one of these six sets.
• G-007 Eighteen set-scope parallel print runs are written from cardboardconnection ALONE without fetching their own tcdb sid: Press Proof Blue #/125 on eight sets, Press Proof Black 1/1 on eight sets, and Press Proof Orange #/99 on Hall Kings and Rookie Kings. The source was calibrated on this release and passed in both directions - 15 of 15 base parallel names, 12 of 12 base parallel print runs, 9 of 9 insert-set card counts, and all five Court Kings parallels confirmed individually on their own sids (Green Flood and Press Proof blank, Blue SN125, Orange SN99, Black SN1) - but calibration is not corroboration of these eighteen specific figures.
• G-008 Fourteen of the sixteen unnumbered set-scope tiers (Green Flood and Press Proof on eight insert sets) had their blank flag column observed on ViewSet.cfm only. The second-endpoint confirmation required before writing an absence was run on two of them - All-Stars Green Flood and The Champ is Here Press Proof, both re-fetched on Checklist.cfm with ?sort=0 and both rendering an empty flag column on every row - and the remaining fourteen inherit that endpoint check rather than carrying their own.
• G-009 On the fifteen base-set parallels the serial token was enumerated across two differently-varied fetches spanning both base blocks (ViewSet.cfm for the head and Checklist.cfm?PageIndex=2 for rows 101 onward, which crosses the 150/151 boundary into the Rated Rookies block). Rows 87-100 and the rows past each page-2 truncation point were never rendered, so a variable run confined to those unobserved rows is not excluded. This is the failure shape recorded in caveats.csv CAV-036's 2026-09-23 extension - one colour sid spanning several base blocks at different runs - and it was tested for at the block boundary on all fifteen tiers, where every tier was uniform.
• G-011 PARALLEL ROSTER SIZES DO NOT MATCH THEIR PARENT SET on six tiers, in both directions, so a parallel's checklist may not be assumed to be its base set's. Larger: Rookie Dominators Signatures Gold 30 against a 29-card base; Signature Series Gold 50 against 48; Significant Signatures Green 100 and Black 99 against 98. Smaller: Swatch Kings Jumbo Prime 24 against 25; Timeless Treasures Materials Signatures Prime 22 against 24; Jersey Series Prime 48 against 50. Which specific cards differ was not enumerated.
• G-012 No pack odds were sourced for any set in this release, and scarcity_tier is UNKNOWN on every set and every parallel. cardboardconnection states only a box configuration ("10 cards per pack, 24 packs per box" with "1 Autograph, 1 Relic Card, 24 Inserts" per hobby box) and no per-set insertion odds.
• G-013 Errors.cfm on the parent sid 153326 returns "No errors or variations have been added for this set yet." under the heading "Errors / Variations". Per reference/caveats.csv CAV-020 that silence is uninformative and is NOT cited as evidence that no variation set exists. The clean, hole-free base run rests instead on a contiguous enumeration of 1-200 across two endpoints, with Gallery.cfm ?PageIndex=21 rendering no tiles. Gallery.cfm strips the b suffix, so it cannot show a variation token directly - but it paginates by card INDEX, and the index-to-number mapping held exactly across every page fetched (page 8 = #71-80, 9 = #81-90, 10 = #91-100, 18 = #171-180, 19 = #181-190, 20 = #191-200, each exactly ten tiles, page 21 empty). An extra b row anywhere in the run would have shifted that mapping and pushed a 201st index onto page 21. That bounds the variation layer at zero EXTRA ROWS; it does not exclude a b row that tcdb has simply never catalogued.
• G-017 rookie_flag=no on cards 1-150 rests on tcdb rendering "RC, RR" on every row 151-200 and on no row 1-150 - a within-sid discriminative split. For rows 1-100 and 171-200 that split was observed on two endpoints (Checklist.cfm and Gallery.cfm); for rows 101-150 it was observed on Checklist.cfm only. No source states in words that those 150 cards are not rookies.

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