The thing is that both of those scenarios are things judges can be expected to understand perfectly well, and have some clue how to approach evaluating the clai

The thing is that both of those scenarios are things judges can be expected to understand perfectly well, and have some clue how to approach evaluating the claim, and also, in both cases, there generally is a reasonable risk that any such attempts leave some form of evidence, deterring people from even trying it. With a smartcard, there is nothing of that sort. It's just a bunch of electronic numbers and "the blackbox says you signed it!", there isn't really anything there to investigate, plus judges can in general be expected to not have even the slightest clue how to evaluate claims about IT security, so more than likely you'll end up with a situation where judges simply accept the government-mandated assumption that what the blackbox says is to be trusted. It's the same fundamental problem as with electronic elections: The actual process is necessarily completely removed from human perception/observation, and therefore ultimately must be trusted blindly if it's not to be rejected outright.