A goniometer would be quite useful for a noise canceller, because of the 360 degree phase rotation available.
I researched this a little.
The goniometer injects no phase shift at all, other than a phase flip.
If field coils were excited in a phase relationship other than zero or 180, then the goniometer could mix the two differing signals in various amplitude relationships, which would rotate phase as the ratio being vector summed changed.
The reason it works with two small loops or Adcock arrays, to electrically rotate those arrays (it really only electrically turns them 180 degrees), is it mixes the signals from those sources in varying ratio. That's all it does.
A pot would do the same thing, if levels could be maintained and if we added a phase flip switch.
This is very easy to model and prove in EZnec. Make two small vertical loops on the same vertical center with the axis of one horizontally in the X direction and the axis of the other in the Y direction. Make one very slightly higher so the wires do not bump into each other.
Now as current ratios are changed from 0 1 to 1 0, we see a 90 degree rotation in null. Now when the goniometer crosses the 1 for the first loop phase will abruptly flip 180 for the zero level loop. Now as level comes up the null moves past and completes another 90 degree swing.
This is what I recalled from the past, although that was before EZnec or computers were thought of.
The technically correct description would be the goniometer changes levels, with a 180 degree flip as one level crosses zero. This can be used to "rotate" two bidirectional antennas, but it really only rotates them 180 degrees. Since that moves one of the two nulls over 180 degrees, we can force a null at any heading (with one opposite).
We could shift phase in a goniometer if we excited the field coils with a phase relationship not zero or 180.