Still brain storming here.
I was fighting a problem a while back and found the wire loop going to resistor door would press in the Potter brumfield relays, with resistor door closed, but I think in my case I think it caused the elevator to stall.

I have also heard of shorts in ribbon cable to logic door causing issues like you describe. But keep in mind, these are long shots.

Its also possible that there are pins touching together on the back side of an IBM relay or ribbon cable terminal block. These terminal blocks are tightly packed and located on the logic door and up on the selector. They are labeled A, B , X, Y, etc.. If someone changed a ribbon cable in the past without proper pin removal and insertion tools , it easy to bend the metal portion enough to touch the neighboring conductor, much less easy to miss wire a single conductor on these terminal blocks. I use a non conductive orange peeler stick to make sure that pins that look a little too close aren't actually touching.

I also wonder if it might be worthwhile to swap optical leveling card with another unit. I wouldn't be concerned about level sensors , because that would just cause failure light to turn on. But perhaps an output component on this card is getting stuck on. These are easy to swap.

Those IBM relays are a punk when you suspect contact troubles. Pulling them out, visually looking through the relay body for debris across NO/NC contacts is time consuming. Pulling and replacing pins is a pain too. Worn NO/C contacts are hard to detect visually. And...got pins? I’ve seen pins x’d over (essentially shorting the contact) and wondered how the unit ran like it did. And you’re right about the relay base socket. Those copper male contacts on the base can fold up upon relay insertion. I’m sure Jim will plug his WizBang tester here if it handles this base.
Would the leveling card you mention fail at only a terminal floor?

If you look at the print, the up directional limit feeds UXA but level up comes in to pick U1, U2 even with UXA de-energized ( Page 3 of the print I’m looking at.) Seems like the car is leveling up but LU never drops. If you can’t catch it in the act with a meter in hand, the optical leveling card might be a good guess. Then again, if it has NEVER kept leveling up through another floor, maybe there is something wrong with the top floor plate. Could the selector carriage stick or bind on level up but the car keep moving?

If you have an optical leveling board, when you get both a LU and a LD signal at the same time, the failure light comes on. This has not been mentioned. Carriage could have a twist and not break the directional. This could happen at the top only, but I'm thinking that the problem would be a lot more obvious. Has anyone actually seen this occur?