That's some impressive code wizardry! I thought the 2-ply search would make it respond to a mate-in-1 threat, but the following game demonstrates otherwise:

1. e4 e5 2. Nf3 Nf6 3. Nxe5 Nxe4 4. Qe2 Nxd2 5. Nc6+ Ne4 6. Qxe4+ Qe7 7. Nxe7 Bxe7 8. Nc3 a5 9. Nd5 a4 10. Qxe7#

9. .., Nc6/O-O/Kf8 would have avoided mate in 1. Maybe this is related to the a2-a4 bug noticed by others?!