Simulating bullet recoil effects on the sear doesn't help much, since the hammer is still forward at that time, as well as the sear itself.
By the time the bolt carrier and bolt have pushed the hammer back out of the way, there's a whole other dynamic going on.
I suspect the design expects the trigger bar to move further forward and then get itself nudged down out of the sear's way so the sear can return back under the hammer's claw.
By limiting the trigger bar's forward movement, I think it stays jammed against the sear, thus allowing the doubling to take place.
It's probably a very unstable geometry which is why it's so random.