I was checking the new Redstone redirection and powering featured in this snapshot and found this bug which is coordinate and direction sensitive.

I'm sure this isn't a dupe of MC-108 as we take that behaviour as an intended feature now.

But as Alugia commented, this might have some relation to the technical aspects mentioned in MC-11193 so this could actually be a dupe of it that had been solved previously and now has come back with these redstone updates.

If you run a Redstone line over not-adjacent pistons:

• Redstone signals disappearing will not retract the pistons under it at pair signal strengths (2,4,6,8,...,14*), unless there is also a powered note block at the side of the piston where the Redstone signal comes from (doesn't work using Redstone lamps).
• In the case of a Redstone line parallel to the Z-axis, if the Z chunk/block is a non-negative one, this only happens with Redstone signals running from south to north (negative Z-direction). And for signal strength 14, piston updating with note blocks only works when they are at the opposite side of where the signal comes from (unless the piston is at the first block of the chunk).

* If at the second last block of the chunk it seems to always retract at signal strength 14.

There are some cases when the pistons don't follow these observations though.

In the attached image you can see the structure I used to test this repeating it in negative and not-negative chunks and X and Z axis (target block and Redstone lamps are for visualizing how far the Redstone signal was reaching).