I mentioned earlier that I was able to reproduce this by turning Bluetooth on and off. Although I cannot see into the innards of OS X, here is my guess at what is happening.
Manipulating the 3DConnexion mouse generates a lot of events that are sent to OS X and then to Rhino. There seem to be some cases (which are unpredictable) where the huge number of 3DConnexion events overwhelm display updates and the display updates do not happen until you let go of the mouse.
I’ve re-prioritized how 3DConnexion mouse events are handled internally and that seems to have fixed the problem for me. This fix will be in the next WIP release. Whether this is a fix for anyone else will need testing with the next release.