Yes, in fact I haven't noticed that this happens in second deployment (thought that it's just second call to save it). Your explanation makes sense; I assume that the fix is only in 5.1.1-SNAPSHOT, right?