The problems aren't really JBoss related. I have very little problems with Wildfly itself. It is a clear improvement over previous versions, I had some problems

The problems aren't really JBoss related. I have very little problems with Wildfly itself. It is a clear improvement over previous versions, I had some problems with v7 which required me to upgrade the JSF modules, otherwise JBoss would suddenly die (in a crash & burn way) with a certain application we had. It's mostly the "memory is not an issue" attitude most java devs have. It's not JBoss specific, we also have some services written in pure Java suffering the same problems - but the majority of the Java apps run in JBoss here. Stress tests in test/qa usually don't bring up the issues I've seen, which sometimes only occur after weeks of running flawlessly with production-grade loads. The 'fix' is then running the critical apps twice behind a load balancer, waiting for one of the 2 instances to die (for logs/memory dumps), automatically restarting them - and hoping they won't go down simultaneously. Truth is - memory management in Java is HARD. When problems pop up, apparently finding the source of it is very difficult. I can't count the times I've heard "we think it should be fixed now", only to be confronted with another alert that the service died again a few weeks later. If it was one service we'd run, it would be fine, but we run quite a few more - resulting in weekly interruptions, tickets, sending of memory dumps to developers, convincing them - again - that it is their problem and not mine, ... So well - I'm sick and tired of Java. The only thing worse to deploy and manage are Ruby apps. But that's a personal opinion.