Bug report filed, http://rt.cpan.org/Ticket/Display.html?id=27944, which includes the work-around of using a long delay, ReadKey(~0/2-1), instead of ReadKey(0), and the simple code change that will likely fix the problem. Thanks.

Too bad it isn't worth filing the bug that ReadKey(0) should have meant a maximum delay of 0 instead of an unbounded delay. :)