havent tried myself, but you might play with try/catch blocks to avoid error messages while being able to catch the exception.
I know this has already been discounted, but I'll reiterate that try/catch only operates via manually thrown exceptions. It does not interact, nor is supposed to interact, with the ArmA error system. I think we've all wished we could catch ArmA errors this way at once time or other ;P
---
Good luck with this, but I'm not entirely sure what advantages this system will have over ArmALib. I would ask for about 1000 new features to come in extensions like this, but sadly, the fact that these systems aren't likely to be universally used any time soon (and having a new one to replace ArmaLib, rather than extending the reasonably popular ArmALib, is only likely to make global acceptance harder), I won't use be keen to use them. As it is, I do add optional functionality to a couple of my systems if ArmALib is available, but I'd be very hard pressed to rely on it. Nevertheless, good luck with this project!
Something I would like to suggest though, would be a modular extension system. If it was easier to extend ArmA with new functionality, rather than relying on Keygetys to do this work, I think that would be a genuinely useful replacement for ArmALib, et al. For example, the easiest way to extend ArmAlib that I can see is to connect to it via a named pipe. Kind of messy and non-multiplexed for a really useful system, I think (i0n0s's editor uses this method, but what happens when 2 systems want to use this pipe at the same time?). I still have a early WIP sitting around somewhere, which is a Ruby backend system that hooks into that pipe to be my own way of extending it (and yes, I was going to multiplex the pipe via sockets). If nothing else you've reminded me I should dig that out again and fix it ;P
Aaaanyway, just some late-night thoughts.