[fadetree]

the skill RNG being streaky is not just a rumor. You can see a repeated pattern even when it's moving along - you'll see 15-20 fails, then a short period where you get one to three skillups, then off on another long fail sequence.
I looked at the core RNG function that the EqEmu code is using, and it's just the usual perfectly good RNG call from the c++ std library using the "Mersenne Twister" algorithm. That is a good quality (i.e., not streaky) generator.
Assuming that the P99 code is probably using the same call, it means that the observed streakiness is not coming from the RNG. I think there almost certainly is some other processing going on underneath the hood that is affecting the results. I am planning to trace what all happens on a skill roll from the time the rng generates its number all the way up to when the skillup result happens to see if I can see any other stuff going on.