If the pairs don't overlap the problem becomes easy, or easier, because now each pair is independent of the previous one and at any point the probability of getting a new pair can be worked out. There's a certain amount of subtlety in counting, but at the very least I can confidently assert that the probability of getting all 100 pairs after generating 200 random numbers (or 100 pairs of random numbers) is equal to 100! /(100^100) (which is pathetically tiny), and that you would only expect to have got all 100 pairs of numbers at least once after having generated 519 such pairs, or 1038 single-digit random numbers (the formula for this one is given by 1+100/99 + 100/98 + 100/97 + ... + 100/2 + 100/1).
What I haven't done, yet, is come up with a general formula for the probability, but the principle is that you can now count "successful" strings more easily, since at the very least they require all 100 two-digit combinations to occur at least once, in any arrangement (100!) and then the remainder can be any arrangement of numbers you so choose, occurring at any point in the string. This counting should be subject to the condition that the probability is zero for a string of less than 100 pairs of digits, and never greater than 1, but always increasing, and so far I've not done the counting properly yet, but never mind. I'll try to think about it a bit more today.