C7: Z-wave reporting devices as failed that are working perfectly fine

I have some simple timers (using WebCoRE) that turn my lights on and off at preset times. Today, one of my Z-wave light switches did not go on as scheduled. While investigating tonight (about 10 hours after the light didn't turn on as planned), I noticed that my Z-wave routing table was as follows (abbreviated to show only the relevant devices):

00 -> 06 100kbps
00 -> 06 -> 08 40kbps

Node 08 is the one that failed to turn on earlier today at 9am. Node 06 was marked as FAILED -- however, node 06 also has a timer that turns it on at 11am, and that timer worked perfectly. So node 06 obviously is not failed. To verify, I tested controlling node 06 from the hub's web interface and I was able to turn it both on and off, after which it switched status from FAILED to OK.

Incidentally, now (12 hours after the missed event, and 2 hours after I last tested node 06), node 06 is marked as NOT_RESPONDING, even though it responds perfectly fine when I control it from the hub's web interface (the status does change to OK after I do that).

Both nodes 06 and 08 are mains powered, so they don't go to sleep.

WebCoRE is not the issue, because I have other WiFi lights that it also controls, and those work flawlessly.

Has anybody had a similar experience? Does anybody have any ideas what my problem might be?

Thanks.

EDIT: I rebooted my hub and the root node changed from 00 to 01... not sure if that means anything...

EDIT2: immediately after that reboot, my hub is now in a state where node 06 is marked as FAILED but I can still control node 08 from the web interface.

EDIT3: I should have mentioned that every single device on my Z-wave network is Z-wave PLUS.

The routing tables update immediately on the device but don't show up on the hub for about 30 minutes if they change just FYI. Node 06 may have a return path communication issue. Maybe a weaker transmit than the hub. Try running a repair on it and see what happens.

My write up was based on the routing table I saw 12 hours after the incident. And as I mentioned, node 06 responded to commands from the hub even though the hub had marked it as failed.