So let’s say we’ve got 28 possible characters (alphabet + period + space), you have a 1/28 chance of getting the first character right. So it follows that a string of length n is (1/28)^n because you have to hit each character correctly in a row. If we can do x guesses a second, and we know on average how many tries it takes on average to get our string right (28^n), we can divide and get an estimate of time. Though we could do it faster, or slower depending on our luck.

With multiple searchers it’s trickier, but we use the probability complement (probability of all possible events must add to 100%) to figure out the chances that our searchers all miss, and subtract that chance by 1. This gives the chance of at least one agent getting it right. Two searchers looks like 1-(27/28)^2 for the first char, and you can follow the same logic for any length string.

Your answer will heavily depend on your assumptions - how fast the computers guess, what they can guess, etc. But searching in parallel would speed things up dramatically. If you had like 100 computers searching simultaneously, 3 or 4 would likely get the first char right every time, giving you a big speed up on the problem.

Yes? The scaling factor is 2^(number of bits).