If the threadpool is terminated for any reason, the appender is unable to recover

[chrismay]

We have an issue where (due to the way that TeamCity build agents launch themselves) the threadpool used by the HTTPCollector will be terminated after the application has started.

In this situation, the Log4jALA appender stops sending logs to Log Analytics and does not recover.

This can be fixed by adding a check in the collect methods that the threadpool is not terminated, and if it is, re-initialising a new threadpool. I have a patch which implements this which I am currently testing, and am happy to offer it as a PR if you think that's an appropriate fix.