New master makes it so enable/disable show as near instantaneous but behind the scenes we still have to do the region close/opens. These can take time.
HBASE-3063 adds into the disable table handler a wait on regions leaving regions in transition but, what if the user did enable/disable/enable/disable then delete. Turns out we don't do this well. It confuses. See the TestThriftServer where we've disabled such a sequence. This issue is about working on making delete work better; e.g. why bother closing stuff if we're going to delete it — just abort the tables regions and save all the flushing, etc., that slows the close-before-delete.