Statistically? With that analogy, you'd design a crypto algorithm that, on average, gets cracked on the desired date to guess the wallet with a few thousand bitcoins. You still have the problem that (1) guessing it right on that desired date, (2) guessing it right today, and (3) guessing it right on a day 42 times the amount of time as the desired date, all have exactly equal probability. You don't get to pick from infinite timelines to make your plot work, you get 1 try, and need to be somewhat sure it'll work on that date. It might not be possible. Unless you time travel and just hand them the information.

No, it's not like that: generate a random Bitcoin address and it's practically guaranteed that it will be empty. But in this case, if the time traveler picks a random key for their secret, it's equally likely that it gets brute forced on the first attempt as it is the trillionth attempt. There's no statistical guarantee except that on average it will take a certain number of attempts.

Sure but the compute difference between decades isn't as dramatic as decrypting vs not decrypting Bitcoin.