With fix for issue 311 [1] the Loggers returned by SubstituteLoggerFactory
would become functional again post initialization. Only the logger calls made
during the initialization phase would be NOOP.

So it might be simpler to just handle proper synchronization around the
LoggerFactory INITIALIZATION_STATE management and continue with current
approach of using the SubstituteLoggerFactory

[1] http://bugzilla.slf4j.org/show_bug.cgi?id=311