Unfortunatelly, no, I didn’t check release history and don’t know, where exactly it was fixed. I just locally changed :5.1.2 to :latest, and it stopped to randomly crash.
By the way, I also found that it was needed to add
KAFKA_REPLICA_FETCH_MAX_BYTES: '1073741824'
KAFKA_MESSAGE_MAX_BYTES: '1073741824'
in environment block in kafka container, to prevent “message too long” errors when sentry upgrade migrates old events.
Although, I still can’t figure out how to fix duplicates errors in postgres (https://github.com/getsentry/sentry/issues/18492) :-/