Z-wave long queue length, init not completing(?)

Longtime OH1 user, my RPi3 died and I decided to just start over with OH2, installed on Ubuntu (virtually, on a Windows box).

Using snapshot apt install.

Almost all my z-wave devices were discovered (reported ~40, there are really ~50, not a big deal, some of the battery ones in particular weren’t discovered).

Added them one by one as things, adding channels, etc. I was able to operate all the devices at that time.

However, since restarting the openhab2 service (I’m using systemctl), nothing’s worked.
I rebooted, habmin claimed the controller device doesn’t exist (it does, but maybe oh2 restarted before the USB aliasing took effect?)
Stop/start service. Devices still aren’t controllable. I see stuff like this in the log (>30 minutes after starting):
2017-04-17 14:34:48.802 [DEBUG] [alization.ZWaveNodeInitStageAdvancer] - NODE 34: Stage INIT_NEIGHBORS. Initialisation retry timer triggered. Increased to 1280000
2017-04-17 14:34:48.802 [DEBUG] [alization.ZWaveNodeInitStageAdvancer] - NODE 34: Node advancer - INIT_NEIGHBORS: queue length(0), free to send(false)
2017-04-17 14:34:48.802 [DEBUG] [alization.ZWaveNodeInitStageAdvancer] - NODE 34: Initialisation retry timer started 1280000
2017-04-17 14:34:48.802 [DEBUG] [alization.ZWaveNodeInitStageAdvancer] - NODE 34: Node advancer: loop - INIT_NEIGHBORS try 2: stageAdvanced(false)
2017-04-17 14:34:48.802 [DEBUG] [alization.ZWaveNodeInitStageAdvancer] - NODE 34: Node advancer: INIT_NEIGHBORS - send RoutingInfo
2017-04-17 14:34:48.802 [DEBUG] [almessage.GetRoutingInfoMessageClass] - NODE 34: Request routing info
2017-04-17 14:34:48.802 [DEBUG] [alization.ZWaveNodeInitStageAdvancer] - NODE 34: Node advancer - queued packet. Queue length is 1
2017-04-17 14:34:48.802 [DEBUG] [ve.internal.protocol.ZWaveController] - Message queued. Queue length = 392. Queue={}

I assume “queue length” is the queue for messages to go out onto the zwave network? Almost 400? Explains why things aren’t working. I expect this is part of the heal, but way more than 30 mins to heal? Sounds wrong.

Do I maybe have a bad device somewhere? Even if I do, shouldn’t OH be smart enough to work around it?
What can I look for in the logs to help debug this? Here’s an example of the z-wave log. (I separated it out from the main openhab2 log as per another thread discussed).

Any suggestions appreciated. @chris perhaps?