My tests so far have been successful. And by "successful", I mean they found an annoying memory allocation bug that would have bitten me on the arse if I'd have just ploughed ahead with the rest of my engine. I'm not sure what's going on in there, but my stress test found that if you disconnect and reconnect slots from signals over and over again that my hacked together memory manager can't cope.
Something is assuredly broken in there, but I'm not sure if it's worth the effort to fix. It's probably better if I replace my hacked together memory manager with something like Boost's pool system, which I found out about with a recent query on the forums. I'm sure Boost handles memory far better than I do, and it will clean up the nasty hacks I had to do to use a std::vector as a memory pool.
This means I've actually gone back a step on my check list; I'm back to working on signals again. Hopefully the new version will pass the unit testing phase.