> since "working" and "non working" status is listed per game, not per driver.
You do get per-game granularity if you use softlists. MESS itself doesn't override the driver's good/bad status with that data, but frontends can and some do show that.
> It would be nice to know which systems are expected to run 99% correct
Those would be the ones where the startup screen doesn't say THIS GAME ISN'T WORKING. (To be fair, a lot of drivers in MESS that do work fine are marked NOT WORKING because the developer is tired of hearing about some specific game that has problems). If you find one that is marked working and nothing runs, it's probably a temporary regression.