We are still using a mix of usernames and uids to identify users (see bug 2754), and getting these out of sync between two HA nodes breaks in a rather nasty way. It's probably not worth actually supporting this use case, but we should at least detect it and fail a bit more gracefully. One main issue seems to be that add_session() updates the uid in the session info. This has the result that any subsequent updates from the other node will no longer match the session. When this happens you see things like this in the log: 2015-02-01 07:49:11 WARNING vsmserver.HA: Tried to transfer session change (update,atril11/vdi2.ensa.es:16) to other node but other node reported HA_NOSUCHSESSION and: 2015-02-02 08:56:01 WARNING vsmserver.HA: Other node told us to remove non-existing session velasco.elisa/vdi2.ensa.es:26 2015-02-02 08:57:49 WARNING vsmserver.HA: Other node told us to remove non-existing session lanza.david/vdi2.ensa.es:20 2015-02-02 08:58:14 WARNING vsmserver.HA: Other node told us to remove non-existing session salcines.mario/vdi1.ensa.es:1 2015-02-02 08:58:23 WARNING vsmserver.HA: Other node told us to remove non-existing session cortes.german/vdi1.ensa.es:9