I've been on a many month long quest trying to track down my Z-Wave message storms. It now looks like the HE Groups and Scenes app is starting to look like the culprit but I'm writing this message to see if other users have seen similar problems.
Quite often I see storms of duplicated log entries from switches and dimmers where they all report in the same information, as if something had issued 5-10 Refresh commands in a row. I don't care about the log. I do care about the WAF plummeting because at this point we are usually waiting for lights to turn on and off for 10+ seconds.
My house has 80-ish Z-Wave devices on three floors. First I thought it was misbehaving Z-Wave devices but after doing a binary search over the course of months I don't think is the case. I only have Z-Wave Plus devices. I also bought a ZWave Toolbox to track this down but nothing looks odd there with routing.
I could never trigger the issue with simple on/off, either physical or digital. I could also not trigger the issue with test Groups that only had devices very close to the Hub.
Creating test groups (or scenes) with devices that have to hop a couple times DO trigger the issue.
This reminded me of various issues I have had with buggy drivers and devices that work when they are close to the Hub but start to fail when they are forced to hop in my largish house. The issue is not the mesh hopping but incorrectly written code that assumes a certain low latency of replies back from devices. I'm starting to suspect that the Scenes and Groups app have some assumptions about timing that have not been tested properly. It's easy so say "DONE!" when you're sitting in a small room debugging code but I would urge driver and app developers, both official and community, to really think about timing and if your code will still work with 1 second latencies between commands and replies.