[xiata]

I know this is naive as hell, but would adding some random sleeps before returning encrypted bits make it better or worse for this particular attack?

[AlyssaRowan]

That kind of "blinding" is entirely ineffective, I'm afraid. It just averages right out.

[AnimalMuppet]

If "sleep" means "sleep until the next timer tick", how can it average out? Especially if the timer is started at the start of the encryption, and all encryptions (at least, for one block) take less time than the timer is set for. That means all times get set to exactly the timer time, and no information reaches the attacker.

What am I missing?

[Natsu]

You always want to remove secret data, not obfuscate it. Think of the difference between those reversible pixel mosaics and a flat black censor bar, for example.

Incidentally, many 'normal' blur functions implemented in software are somewhat reversible.

[TheLoneWolfling]

It would mitigate it slightly, but ultimately will be ineffectual - they can simple take more measurements.