Fixed
- Crash loops on containers without a healthcheck could go unnoticed (#2, @NotRetarded). Docksentry watches a container's RestartCount to catch crash loops — no healthcheck needed — but the detector only fired if it caught the container in the
runningstate at sample time. A container that crashes instantly on startup (e.g. a VPN container failing on an unsupported kernel) sits in restart-backoff almost continuously, so a once-a-minute check rarely saw itrunningand the loop slipped past. Detection now keys off the RestartCount increase itself, regardless of the sampled state, so a container thrashing on startup alerts even with no healthcheck. The 30-minute per-container cooldown keeps a fast loop from spamming.