The default value of executor.max.threads is 100 which is fine for average sized instances, but probably too small for very large instances. This limit is often the first one to be hit on instances with very high (especially SSH clone) hosting workload. The value should be increased, preferably scaling with instance size.
The logs below most likely tie back to this issue:
2015-07-12 01:03:10,900 ERROR [threadpool:thread-12] c.a.s.i.c.StateTransferringExecutor Error while processing asynchronous task
java.lang.IllegalStateException: Interrupted while waiting for delegate SCM request
at com.atlassian.stash.scm.cache.ssh.UploadPackProxy.startDelegateRequest(UploadPackProxy.java:365) ~[na:na]
at com.atlassian.stash.scm.cache.ssh.UploadPackProxy.streamRefs(UploadPackProxy.java:230) ~[na:na]
at com.atlassian.stash.scm.cache.ssh.CachingSshUploadPackRequest$RefsContentProvider$1.perform(CachingSshUploadPackRequest.java:423) ~[na:na]