I don't think this really helps you. But when checking the JBoss support policy (Red Hat JBoss Middleware and Red Hat JBoss Operations Network Product Update and Support Policy - Red Hat Customer Porta…) you may realize that the
Extended Life Support for 4.2 had ended in June 2015. So this is quite a old version and I'd suggest that you plan for an upgrade.
But id might be useful to post the server.log extract at the point of failure. Maybe somebody is able to help you anyway.
yeah, 4.2 is quite old
But the problem you see seems not related to JBoss.
The system might be overloaded either by increase the load and need more heap or slow down some resources where each transaction take longer block resources and memory and run into GC issues.
You need to tune the GC / heap. First of all I would check the gc.log (there are some helpful tools like this chewiebug/GCViewer · GitHub) another tool included in the JVM is jstat where you can see some memory statistic and GC activity.