As scary as this can read, it is actually a very minor bug.
Instead of following a 1:(1*100) sampling, the events may follow a 1:(100*100) sampling. However this is not messing with the stats because the applied sampling is sent with each event anyway.
So this is no big deal, but still should be fixed.