We're using logback 1.1.3; which as you note is not aware of the event replay capability. We're OK with losing a few log events which occur during initialization, and we've been living with that in earlier slf4j releases.
So the logback version may be part of the problem, but there also seems to be a concurrency element because this only occurred on a small percentage of our servers.
So far I can't get the eventQueue in a simple unit test using the same versions, but my hunch is that in some condition replaySingleEvent()
>isDelegateNull()
>getLogger(loggerName) returns a reference to a substitute logger instead of a 'real' logger so the substitution remains in place.
|