Because of overloading RetryException, TripleWriter and BlobWriter will now retry endlessly if the connection pool is 100% busy. This is likely not a large problem unless the application is really overusing database connections, but could result in very long execution times when what is needed is really an exception indicating that the connection pool is oversubscribed.

There are two solutions:

1. quit overloading RetryException and have connection-fetching code throw some other exception when no connection is available

2. put a retry limit on the retry loops that currently handle RetryExceptions resulting from deadlocks.