LabelRelay backed off and won't reconnect to self-hosted Ozone labeler

[clarabellexyz]

My self-hosted Ozone labeler's WebSocket is working correctly (confirmed via wsdump both directly and through Caddy), but the LabelRelay has stopped retrying after a series of earlier 502 errors caused by a Docker DNS issue that has since been fixed.

Labeler DID: did:plc:roewhjzra5xtuliye7qf2ghy
Ozone endpoint: ozone.booksky.club
Relay's last requested cursor: 102
Current label sequence: 103
Last relay attempt in Caddy logs: March 25 00:09:49 UTC (502 error, connection reset)
WebSocket confirmed working: wsdump wss://ozone.booksky.club/xrpc/com.atproto.label.subscribeLabels?cursor=102 returns labels successfully

The relay was hitting 502s for several hours due to a Docker networking issue (now resolved — Caddy moved to host). It appears to have backed off and hasn't retried in 19+ hours. Is it possible to nudge the relay to retry our labeler? TIA. 🦋

[clarabellexyz]

Resolved. The fix was waiting a week and then restarting the Ozone instance (docker compose restart ozone). I had already done this on the 25th but after restarting again today, the relay reconnected on its next retry after the restart.
For anyone else hitting this: the restart seems to trigger the relay to attempt a fresh connection. The catch is that Ozone takes a few seconds to boot, so if the relay retries too quickly it may get 502s during startup — but it came back on a subsequent retry and labels were appearing within an hour or two of starting this process today.