A restart of NWare solves most "exception-***" errors. I usually see this after I have caused the network card (the NIC in your computer) to stop talking "cleanly" to NWare, either by changing an IP address, changing network VLANs, etc. Everything else will seem to work, but when a deployed is tried, an exception occurs (too much data flying around). NIC communiction of NWare is much more stable and robust than Pandad, but the communication path to the NioNode can still be broken just enough to cause deploy problems.
This will solve the above problem. First, shutdown all instances of NWare, wait several seconds (to make sure NWare is clear of all memory), and then re-start NWare (make sure Pandad can see your NioNode in the Remote Log, to make sure Pandad is working correctly). Now you should be able to do a clean Deploy. This usually happens when I change IP address because I need to check the CobraNet network, or because I need to get on an outside network to check email, and then go back to the Control Network to deploy a change. Most of the time it is not a problem, especially if I "disconnect" from the Nions before changing networks. But sometimes...
Another possiblity is that the Compiler did not see a change that was made to the Project that should have required a re-compile. The solution to this is to press and hold the "Shift" key and then start the Compile by your favorite means (click the running man, press F9, file/click Deploy). You will know if you have made a change that should have required a new compile, but NWare went straight to deploy instead of compile and deploy.
If I do the restart NWare step, I will normally do the shift and compile step (by then I'm too frustrated to NOT take the time to "force" a compile).
However, your comment about "a unit that theorectically_should_be ignored" is something I've never seen before.
Make it intuitive, never leave them guessing.