Currently if a error handling callback occurs during logging, and the callback tries to log, the process will deadlock. I know this can happen in the crtDebugCallback callback, and I worry that it's possible that it would occur in stack tracing or similar.

We should investigate the use recursive locks in our logv2 locking front ends or back ends.

Alternatively we could think about an how to create and utilize an alternative lockless logging API for use in such callbacks.