pathbias_get_scale_factor() is using the wrong default and also causes rounding errors
pathbias_get_scale_factor() returns DFLT_PATH_BIAS_SCALE_THRESHOLD instead of DFLT_PATH_BIAS_SCALE_FACTOR as the default value.
This mixup means that every time we hit DFLT_PATH_BIAS_SCALE_THRESHOLD worth of path bias data, we basically totally reset our counts.. Woops.
Also, my analysis from #6135 (moved) makes me think I did this whole scaling thing wrong. It should be fixed-point decimal value that we multiply with instead of a divisor. This will allow us more flexibility, but it will also allow us to easily approximate a moving average, which is I think probably what we really want. See #6135 (moved) and links for details.