This would be great, except that tvheadend keeps overwriting my corrections with incorrect info.
FWIW, it appears my initial suspicion is probably correct, because what appears to be happening is that channel information is nearly always associated with the next
higher mux, meaning that tvheadend thinks the tuner has changed frequencies before it actually has. w_scan doesn't have this problem and correctly output all the proper muxes on the first try, presumably because it scans more slowly to tolerate sluggish tuners.