A critical regression has been discovered while investigating the ability to delay the popup of the web browser until the server is finished deploying.

A commit was pushed in on July 7th, 59 days ago, reversing the logic that was used to determine if the deployment scanner should be stopped or not. The original (and intended) behaviour was that if the server is started, and a JMX extension is provided, to disable the scanner during publish events.

The offending commit is here:
http://source.jboss.org/changelog/JBossTools?cs=32709

You can see that the logic was removed from one class to another. LocalJBossBehaviorDelegate had the following code:
protected boolean shouldSuspendScanner()

{ return getActualBehavior().shouldSuspendScanner() && ExtensionManager.getDefault().getJMXRunner() != null; }

The code in the "actual behaviour" had the following:
if( getServer().getServerState() != IServer.STATE_STARTED)
return false;

The code in LocalJBossBehaviourDelegate was replaced with:
protected boolean shouldSuspendScanner()

{ return getServer().getServerState() != IServer.STATE_STARTED && ExtensionManager.getDefault().getJMXRunner() != null; }

You can see that the original code ensured that if the server state was NOT started, it would NOT suspend JMX... ie, that the server being started is a requirement to suspend JMX deployment scanner. In the new code, you can see that it will TRY to suspend if the server is NOT started, and that if the server IS started, it WILL NOT try to suspend.

It is extremely obvious that no testing for this usecase was performed during the time period around this patch.