No, I don't.

 

It is fairly easy to reproduce. Just add a JDBC-driver to an agent, write a little code to call a database, schedule the agent to run every x minutes and wait.... It will crash within a fairly short number of calls (depending on the environment and code, expect 5-10 runs). 

 

I have spoken to some of the developers on this at LotusPhere the last couple of years. It seems they need to re-write much of the way the JVM was implemented for agents and webservices to fix this. A much better approach going forward would be to look at OSGi to do some of the same things - and obviously find a solution for deployment. But it seems someone is working on/considering this.

/John