> Our full dump shows no deadlocks and nothing waiting
> for a monitor (We visually analyzed and ran through a Thread Dump
> Analyzer to confirm). The thread in question is runnable (no java
> monitor deadlock) and just seems like it is waiting for a reply from
> the database after execution of the sql statement via a socket read.

Reason for this can be also a deadlock on the server. You can check if
this is a case by reducing the deadlock timeout on server via its
config. Alternatively you can play with transaction parameters
(WAIT/NO_WAIT transactions).

The only thing that does not really fit the picture is the INSERT, which
under normal conditions should not deadlock. But, if I am not mistaken,
the deadlock is still possible even for inserts in rare cases (IIRC, it
is related to allocating new pages or index updating...).

Roman