[brucehoult]

It's not like it's hard to find a harmless 3 cycle instruction.

For example Store A/X/Y to any unused/scratch Zero Page location.

Or condition branch to the next instruction using a condition known to be true .. BCC/BCS is often the most convenient. Or just JMP to the next instruction, though it's 3 bytes of code instead of 2.

INC/DEC of a scratch Zero Page location is useful for a 5 cycle delay in 2 bytes of code if you don't mind changing NZ flags, but the store/branch/jump alternatives don't even do that.