If there are 
org.infinispan.commons.util.concurrent.CacheBackpressureFullException
it is very likely that threads are hanging, other possibility is the server is overloaded.

To track that it is important to have a thread dump or a series of it automatically for post-mortem analysis.
To prevent from overloading the logs it should trigger a few in a row (if necessary) and stop for a longer time. Also it might be an idea to not log the complete stacktrace in this case.