Consider rewriting test-gainnode.html using audit.js.  If the test fails it prints out the first point where the difference between the actual and expected result exceeds the threshold.  Which is ok, except if you're changing the test and need to update the threshold. You get to update the threshold basically one element at a time and also manually computing the threshold after each failure.  This is really horrible.

Compare how this would work if we used audit and should().beCloseToArray().  The failed output would show you what the actual max absolute (and relative) error for the test would be so you can just look and set the threshold after running the test just once.

Also, the test doesn't allow any way of updating the expected wav file from the actual output.  That's pretty terrible too.