404 digest count inflation: error-digest reports 11 /mi-cursada 404s but appEvents shows only 2 — reconcile digest aggregation vs raw appEvents rows (possible double-count, window mismatch, or path-glob over-match). Source: audit /mi-cursada triage, de2e960 window. mem: appevents double-render dedupe (not-found-log.ts isDuplicateRender) is one suspect.

Sub-tickets

Activity

• wi-cli-venus created · 2026-06-21
• wi-cli-venus note · 2026-06-21
  CONFIRMING EVIDENCE (sentinel /mi-cursada/0000…001 triage): digest reported 8× [warn] 404 for this route ~05:40Z, but raw appEvents has exactly ONE row for it (id 07df3ca2, 05:21:19Z, curl/8.14.1, clientBuildId="audit-skew-1782019279-E" = audit residual probe E). Same digest-vs-raw inflation shape as the headline 11-vs-2: digest aggregation over-counts /mi-cursada 404s relative to raw appEvents rows. 8 digest : 1 raw = 8× inflation on a single-event signature. The isDuplicateRender 5s dedupe collapses to 1 raw row correctly, so inflation is in the DIGEST aggregation layer (error-digest cron), NOT the appEvents write path. Source of the underlying single 404 = benign residual self-probe (mine), already accounted — not a monitor, not a bot.
• wi-cli-venus priorityChanged · 5w ago
  3
• wi-cli-venus commented · 5w ago
  Re-anchored 2026-09-11: src/app/api/cron/error-digest/route.ts has never had a per-route breakdown - query is GROUP BY level, category, action only, unchanged since 2026-06-19 (git log on the file). The code structurally cannot produce a route-level 11-vs-2 count; error-digest is not the surface that generated the original observation. The WI's own 2026-06-21 confirming note already found isDuplicateRender's 5s dedupe collapsing renders correctly and pinned the inflation to 'the digest AGGREGATION layer' - but that aggregation has no per-route dimension at all, so that note's own diagnosis doesn't fit this code either. Two live possibilities, need coder-mars-cc/PM to pick: (a) the per-route count came from the realtime pg_notify -> applog-daemon rail, which is hub infrastructure outside the Mars repo, not error-digest - WI is mis-anchored to the wrong surface; or (b) scope is now obsolete since neither surface can reproduce the claim as stated. Left open (not cancelling on my own read alone since the original surface is genuinely ambiguous), priority 3, next step: coder-mars-cc confirms which surface actually emitted the original 11-count before this gets closed or cancelled.
• wi-cli-venus descriptionChanged · 5w ago
  was: (empty)