Argh, I uploaded the last beta build one day before I made this fix in my sources. So here's a new beta build which should now really fix the problem:
http://madshi.net/madCollectionBeta.exe (installer 2.8.4.11)
I don't understand why it would be a critical / show stopper error, though. Leak reporting is not supposed to be turned on all the time, because it eats lots of RAM and slows everything down. It's intended to be turned on only on the development PC, or maybe sometimes for a very specific test on a customer PC. Of course madExcept itself should (usually) be always active. But leak reporting is a completely separate part.
Actually, unless you manually distribute madExcept32.dll with your EXE file, leak reporting will be automatically disabled on your users' PCs. Leak reporting requires madExcept32.dll to be present. On your development PC madExcept automatically finds madExcept32.dll if madCollection is properly installed. On your end users' PCs usually madExcept32.dll is not present. So by default leak reporting should be disabled on your end users' PCs, even if you have it enabled on your development PC.
If you get random errors, they are very VERY unlikely to be caused by this "wrong detected resource leak". All this little bug does is report a leak which isn't really a leak. But this should not result in any random errors. Furthermore, as described above, leak reporting shouldn't be active on your end users' PCs, anyway. So the reason for those random errors is probably something completely different.
If your app crashes and the code-break-point is inside DevEx/TMS/LMD code, then why do you think this has anything to do with leak reporting? Does this problem go away if you disable leak reporting? What is the exact madExcept crash report you get from this situation?