As written in https://issues.jboss.org/browse/WFLY-3198?focusedCommentId=12958470&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-12958470

We should look into using the management API notification stuff Jeff Mesnil has done/is doing. The scanner subsystem can register listeners and the deployment ops can emit events. This can be used to keep the scanner aware of changes to the deployments part of the management model independent of the periodic scans.

This would close a gap that exists where the scan could be configured to run infrequently (say 10 mins), the .undeployed app gets deployed via console, and then before the scanner runs again and cleans out the .undeployed, the server is reloaded/restarted.