There is a non-zero chance of a java.lang.Error stopping subsequent executions of the RetryRunner task.
CassandraHostRetryService.verifyConnection should catch and log Throwable instead of Exception. Also, it would be nice if that log message included the affected host.
In the end, the error turned out not to be originating in verifyConnection, however verifyConnection should probably still catch Throwable instead of Exception in order to be able to not interrupt checks for other hosts.
Pull request to follow...
RetryRunner Dies If ThriftCluster Was Not Created Via HFactory. Fixes…
… issues #423 and #424.