Document that Gitea reports combined state as `failure` when ALL
individual status contexts have `state: null`. This causes spurious
`[main-red]` issue filings from the watchdog — first seen on
issue #481 (2026-05-11).
Also removes the now-redundant "fetch-depth: 0" section (duplicate
of the "Large repo causes fetch timeout" section).
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
SRE review of PR #457 flagged two factual errors that were not
addressed before merge. Applying corrections directly per SRE
mandate: no manual production changes without config-as-code.
Corrections:
1. Remove "git fetch --depth=1 times out" — shallow fetch succeeds
in ~16s per PR #441 detect-changes evidence. Only fetch-depth:0
and git clone time out due to ~75MB repo history size.
2. Rewrite "runner cannot reach git remote" to accurately state:
runner CAN reach the remote; fetching full compressed history
exceeds the ~15s network timeout window. Repo-size constraint,
not network isolation.
3. Updated diagnosis snippet and verification section to match.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
