[iamtoruk]

Summary

• Codex now moves cold rollouts to rollout-*.jsonl.zst, which CodeBurn skipped, so most Codex history disappeared from reports. Every Codex reader (discovery, parser, fork lookup, codex-tps, context view) now decodes them.
• A session present in both forms counts once, from the plain file. Compressed files are parsed whole and never resumed by byte offset. A damaged .zst prints a visible warning; usage after the damaged part is not counted.
• On Node below 22.15 (no node:zlib zstd), compressed rollouts are skipped with one notice. Results there match current main, and the daily cache stays at version 75, so a later Node upgrade adopts it and re-derives the settled days.
• With zstd, DAILY_CACHE_VERSION goes to 76 so settled days pick up the recovered usage.

Testing / Verification

• I have tested this locally against real data (not just unit tests)
• npm test passes
• npm run build succeeds
• Not a UI change

Real ~/.codex with 1,436 compressed and 6 plain rollouts, 2026-01-01 to 2026-10-25:

	Sessions	Calls	Cost
main	3	54	$0.16
this branch	1,358	51,564	$4,842.39

• The same files decompressed and read by main give a byte-identical export.
• With zstd hidden from node:zlib, the export matches main exactly, prints the notice once and writes a v75 cache; the next run with zstd adopts it and reaches $4,842.39.
• A daily cache written after compression goes from $0.16 to $4,842.39 on the first run. A cache written before compression keeps its totals, with 0 days changed.
• Eight plain/compressed transitions on the same session set (compress, delete, both forms present, back to plain, fresh cache each time) give identical totals.