Ah. It's an issue with Piston Quasi-connectivity I think (see Quasi.png). MC-108 Technically "works as intended", but Grum is working on it because he also thinks it's a bug.
Edit: However I could still retract the piston with my setup. So maybe MC-11193
migrated
Definitely MC-11193. The torch is, for some reason, staying powered just after the wire is depowered, so when the piston is updated by the wire, it is still powered.