With help from Ralf on this comment I was able to

1. Disable the memory by no-longer grounding PIN 32, and sometimes forcing CE high on the SRAM chip.
2. Boot a workbench disk, and then break it on startup.
3. Insert a disk containing addmem from ykick and memtest.
4. Re-enable the commodore amiga A501 replacement prototype.
5. Run “addmem $c00000 $c7fffe”
6. Then, run memtest.

As you can see (and this is true over a large number of errors), databit 10 is always stuck high.

I’ve got to figure out why this is……..this doesn’t go through any latches or logic on my board.