I got ahold of version three of Rush Hour, which contains all Lucien's fixes, and is considered the final code.

A disk image was made of the compiled, unmodified code. Seems to work just fine. Now I can start removing his workarounds and restore the code he originally meant to write.

First step, removing the unneeded "mov r1, r1" from vdp_copy_from_sys.
Works, but the cursor is corrupted after moving a piece. Hmm.

Dump of original code:

00000000 :
0: c0 41 mov r1, r1
2: c0 41 mov r1, r1
4: c0 41 mov r1, r1
6: c0 41 mov r1, r1
8: a0 c2 a r2, r3
a: c1 01 mov r1, r4
c: 06 c4 swpb r4
e: d8 04 8c 02 movb r4, @>8C02
12: 02 61 40 00 ori r1, >4000
16: d8 01 8c 02 movb r1, @>8C02
1a: 80 c2 c r2, r3
1c: 14 00 jhe 0

This seems to be inlined all over the place, so I need to look closely at all references.