Did you modify the software in any way?
I can't remember having issues to submit the first hit of a work item solution many many years ago when vanitypool was rather new and there were plenty of not terribly hard work items available (GPUs back then weren't as powerful as today's and I only had a barely middle-class one).
It would be against all odds that I was purely lucky that my first found solution to a work item was just one that worked out fine, if there were detected "solutions" that won't spit out the correct public address prefix after combining the keys.
Why would there be something like "1 in 6" chance? That's nonsense in my opinion as I don't see any mathematical proof for such ambiguity. This sounds much like some strange bug in the software.
And I also can't remember any discussion about such an issue in this thread earlier (not that I've followed the many pages of it to the word).
I use VanitySearch for searching, with a special search scheme, for speeding up. This problem arose only when searching for a prefix with a public key, when searching for a prefix without it, 100% match of results. I also thought that when using a public key, the match should also be 100%, but it turns out that no, for what reason I can not answer exactly, maybe because elliptic curves are used for calculations or this is a bug in the program or memory without error correction, it is not clear yet why this happens, too little data and it takes months to accumulate it. Maybe there were no problems before because the prefixes were not complicated and it was possible to quickly find another option.
The only question I really have is why does this happen with LEGACY wallets! Why do legacy wallets have child keys?
There are more people who would probably know what's going on here if they see the post. I didn't get into VS until 2019 or so.
You were right vanitygen gives 100% match when searching for a prefix with a public key, but it works almost 20 times slower because it does not support Nvidia cards!