Summary: a2audit (1.06 and 1.07) is attempting to do the SOFTSWITCH TESTS for this empty slot.
- specifically it does
lda ($fe),y when ALTZP=1, ie. the CPU reads the floating-bus for zero-page values $FE and $FF
Here's AppleWin: (-aux empty)

Couple of thoughts:
- Move the ALTZP test to be 2nd in the
.writeswitches data table, and skip both RAMRD and ALTZP tests for 64K machines
- Or change the test code so it doesn't use
lda (.resetloc),y etc. and instead uses self-modified lda abs16 addressing mode (to avoid zero-page)
Perhaps (2), as testing ALTZP even for an empty aux slot is still a valid test for the IIe.