.... I would have loved to not run into the first problems already when running the setup. Two things to watch out for so far:

1. The setup seems to change passwords on the proxy accounts, so if you upgrade while the engine is still running and has events pending, you'll get lots of "authentication failed" messages at the console and it won't unload. Instead it retries to finish the current process forever (or at least longer than my patience allowed for). Since luckily-soon-to-be-extinct Netware somehow made it into 21st century without a task manager, all you can do to get the new engine version loaded in this case is reset the server.
Do not forget to unload the engine before you run the setup!

2. Updating event and sentinel servers through nsmadmin does not always work properly over WAN links. In our case updating 8 out of 10 remote sentinel servers (connected over 2-4 MBit lines) did not succeed and trying to load fsfevent.nlm on the console results in [BAD FORMAT]. No wonder, the file did not get copied properly and has a different checksum than the one on the engine server. Again, unload all NLMs and copy them manually to the event/sentinel servers instead. This is faster than using nsmadmin anyway, only one question arises: if NSM can't even copy it's own NLMs properly, can we trust it to handle user files correctly?