Summary

• refresh the GitHub rate-limit snapshot on each scan instead of letting cached daemon pause state short-circuit the scan loop
• keep using the cached pause snapshot only as a fallback when the live rate-limit query fails
• add a regression test for the stale cached core_remaining=0 case where the live GitHub quota has already recovered

Root cause

The daemon cached an in-memory GitHub low-quota pause and treated it as authoritative on later scans. If that cached state drifted from the live GitHub quota, the daemon could keep pausing scans even though vigilante status showed healthy quota from a fresh rate-limit call.

Validation

• go test ./internal/app ./internal/github