I don't follow this. The trim is not for setting DC offset, it's for setting the quiescent current. DC ofset should be minimal because of C3, which makes sure all DC feedback goes to the inverting input of the long tailed pair.
C3 won't see any (significant) DC unless in fault conditions. By your argument, you could say the input cap will also be destroyed, because it also sees changing polaraties of the input signal.
I can't really comment on the circuit behaviour you describe here, because it's somewhat beyond me, but I can tell you that this amp is very stable. Not only is it my experience (I have already made one, but now I'm redesigning the PCB), but it has been built by many others. It's predecessor
The lack of a current limiter is intentional BTW. The author hates the reduced (near) clipping performance caused by such a thing. This is an amp for DIY, not mass production, so a short circuit protection is not necessary. And, as you can read in the forum thread I linked to, the overload situation certainly does not happen according to one of the posters.
Also, the ESP website is not just made by a random electronics tinkerer. Rod Elliott really knows what he is doing, and has a lot of experience in audio electronics. There are over a 100 projects on that website he designed. And, they all work as stated in the articles.