Weirdest? Some memory trashing which just so happened to replace a branch-if-equal with a branch-if-not-equal, and that code just happened to be in a multiplic

Weirdest? Some memory trashing which just so happened to replace a branch-if-equal with a branch-if-not-equal, and that code just happened to be in a multiplication subroutine (no hardware multiplication) causing multiplication to always give a negative result, post-trashing. This was showing up in some compiler-generated offset multiplications which were optimized out unless it was a debug build, long after the memory trashing took place.