I was able to make 3 of the AppleTi cards and tested them today with pneubauer's Kansasfest 2016 presentation disk. Everything works great!
I did find that they are incompatible with a ZipChip. Initially, the Ampersand driver would crash, or sometimes it would load but the PRESENT or GAME programs would crash. I tried moving the card to different slots with varying success, but the programs never worked correctly. I tried disabling the ZipChip via their utilities, and making sure the slot speed was set to normal instead of fast, but nothing worked. So I swapped out the ZipChip for a normal 65c02 and everything worked like a champ in any slot I moved it to.
I've tested in a IIgs at 2.8 MHz. Maybe that isn't fast enough to show the problem, or maybe the issue is unrelated to speed. When AmpArcade crashes, what happens? Does it break to the monitor? If so, what's the program counter, and what data is at that address?
AmpArcade tends to crash if it's unable to detect the card. Maybe there's a bug in the detection logic. Also, AmpArcade scans slots to find the card, and the scan process involves writing to whatever cards are in your slots. That could "mess up" other cards. What happens if you move the AppleTi to slot 7 (the first slot in the scan sequence)?
On close examination, that sprite demo disk has some odd imagery. One sprite is obviously flipping you off, one is dropping a deuce, one is seemingly popping pills. The demo was seemingly written by hippies. Now get off of my lawn!
Many of the AmpArcade and SuperSprite demos do not work with the AppleTi. Those vintage cards include sound and video mixing hardware. When the software attempts to write to that non-existent hardware, you end up with random writes to the F18A registers. That has various odd effects. You might be seeing these effects -- you'd have to remove sound and video mixing hardware accesses from the demo to confirm this theory. Or, if the demos use AmpArcade rather than direct POKEs, we could disable the unsupported commands in AmpArcade.
I would be curious to know if the problem is the software or something else. The F18A should easily support the faster CPU. Any possibility that you will be writing some of your own software to use the F18A?
I'll investigate the software side. I do plan to extend AmpArcade with F18A-specific features, but, like all hobby projects, that might take awhile.