The issue with testing this properly at the moment is that the event log is too barren to show us whether it's happening or not. All we can see is the end result - even with advantage, we see one singular outcome in the event log.
I'm inclined to think that it is working in some specific cases, and not in others.
For example, I did a test of 400 rolls on the checks to read the magic book, for another test (Testing roll bias). The character was a halfling and I never saw a 1 rolled in all that time...
However, that same halfling has had 1 show up on the die face in other ability checks elsewhere in the game.
That halfling has also had advantage attack rolls claim to have been rolled as 1s, which, for a halfling would require 4 natural 1s in a row... that I've had that occur more than once in my play time is distinctly concerning... But, without being able to see what it claims the actual die rolls are, all together, we simply cannot check if it's working or not; not conclusively.
