VersionEye

Reliability release, prompted by a real Modrinth API outage that took down 21 of 28 checks on one server.

Fixed

• Failed checks are no longer counted as up to date. Summaries now exclude them and list exactly which plugins could not be checked, both in scheduled checks and in /updatecheck.
• An API outage no longer wipes what VersionEye already knows. Updates found before the outage stay available to /updatecheck download and /updatecheck ignore, and join notifications keep working.

Changed

• Lookups are retried once, two seconds apart, after a 5xx response or network error, so a brief API blip no longer fails the whole check. The Discord webhook is never retried, so it cannot announce the same update twice.
• After 4 consecutive failures to the same API host, the remaining checks against it are skipped for that run with a single "appears to be down" warning instead of a long series of timeouts and per-plugin warnings. The next check starts fresh.
• Request timeout reduced from 15 to 10 seconds; combined with the retry this keeps worst-case check time about the same while handling slow responses better.