Well, from my fuzzy ancient memories....
A v4 server hiccup, the details of which I've long since forgotten about, caused some LL results to be lost -- only recoverable from a report that showed the masked bottom 8 bits.
At the time, I figured it was best to add the results to the database without valid bottom bits so that the exponents would not be handed out as first-time tests and the user received proper CPU credit.
In looking the last pre-v5 database, it looks like there were 4 such cases: 18050569,19857121,33238477,33254747
|