I find it fun that some of the stolen tokens appear to have been donated back to a Bitcoin advocate.

Too bad you didn't try it with much higher degrees.

The "infamous" reference generators from NIST 800-22 included linear, quadratic and cubic congruential generators only. A potentially vulnerable implementation that may have used this document as a reference would probably have only gone up to the cubic case. So I think it's unlikely that someone used a recurrence equation of higher degrees. But you never know. Also, the higher the degree, the more resources the attack will require. So, we opted for a balanced cost/benefit approach.

Yeah, at least to try on that one signature with 3 million wallets on it. If it doesn't work with 30 or 100, most likely it means they're using a valid number generator, but if not, damn. That's a lot of bitcoins