During a clean tomcat startup (tomcat was stopped before) and initialization of logback we observed that an AsyncAppender may come up without worker thread resulting in a deadlock of the application, because the queue will never be emptied.

We use a lot of AsyncAppenders in our setup, and this Error seldomly happens, so this might be hard to trigger. I suspect a Race condition somewhere?

Here is some logback output with debug turned on:
You see the Worker Thread dies immediately after initialization and nonetheless the initializian continues and the Appender gets added to the ROOT logger.

12:07:07,295 |-INFO in ch.qos.logback.core.joran.action.AppenderAction - About to instantiate appender of type [ch.qos.logback.classic.AsyncAppender]
12:07:07,295 |-INFO in ch.qos.logback.core.joran.action.AppenderAction - Naming appender as [XXX-log-error]
12:07:07,295 |-INFO in ch.qos.logback.core.joran.action.AppenderRefAction - Attaching appender named [XXX-sync-log-error] to ch.qos.logback.classic.AsyncAppender[XXX-log-error]
12:07:07,295 |-INFO in ch.qos.logback.classic.AsyncAppender[XXX-log-error] - Attaching appender named [XXX-sync-log-error] to AsyncAppender.
12:07:07,295 |-INFO in ch.qos.logback.classic.AsyncAppender[XXX-log-error] - Setting discardingThreshold to 51
12:07:07,296 |-INFO in ch.qos.logback.classic.AsyncAppender[XXX-log-error] - Worker thread will flush remaining events before exiting.

...
a lot of successfull appender initialization cut out
...

12:07:07,338 |-INFO in ch.qos.logback.core.joran.action.AppenderRefAction - Attaching appender named [XXX-log-error] to Logger[ROOT]