EY tripped, recovered, I started realigning Y arm, and at some stage when the arm transmission started to become decent the transmission number started jumping back and forth.

At first I was tempted to think of this as a HOM-related fluke, but when I looked at the time series, it looks as if some incorrect bits are flipped at certain number (though the data is already decimated to 2kHz and the decimation filter should make things somewhat fuzzy).

I'll connect a function generator or something to this channel and see what's going on on Monday.