Thanks for the cautionary note!

I wonder if the inconsistency might cause the sort to continue making swaps indefinitely, thus causing the failure to terminate that you mentioned?

Of course, my simple test case did just fine, but that probably was dumb luck.