The issue was an errant reset signal presented to the clock counters, which had
an active-high reset signal instead of the normally-used active-low reset. This
has been fixed and the test logic now works as expected.