> I wonder if a pattern like this would also be useful as a "jitter check" --- which bits change and the statistics of their values over many reads could show how closely aligned the drive mechanism is.

Maybe this type of global synchronisation was built into the FDC? You don't need fuzzy bits to detect this, at the most basic level the PLL keeps the local stream of bitcells in sync in the current sector... the deltas (as a side effect of the sync) could be used to calculate an average and change the global expected timing of the window - as it collects more deltas it would converging on the global difference between the physical characteristics of the drive and the disc timings (i.e ultimately the difference between the drive that wrote it and the drive reading it).

If the FDC failed to read a sector due to PLL synchronisation (would that result in an invalid MFM encoding sequence?), then it could attempt to read other sectors to gather more delta samples and obtain more accurate expected timings before trying again.

Since the fuzzy bit trick still produced valid MFM sequences i suppose this type of functionality would not get the drive stuck in that case.