Possible solve...
I may be speaking too early on this but it's looking promising.
I've been super focused on what could cause the filament to start a phase transition
in the wrong location and I THINK I've got it.
The short of it is that the PTFE tube connecting the extruder to the hotend may have been a hair or two short of seating solidly against the heatbreak.
Long version...
The way this probably came about is, because the PTFE collet on the bottom of the extruder is not visible from above (I have a Railcore 300ZLT so it's pretty tall and sits on the floor), I'd gotten in the habit of leaving the collet on the extruder and removing the one on the heatsink coupler and just pulling the extruder and tube out in a single step. When reassembling the extruder after cleaning it, I probably seated the tubing as far into the extruder as I could, leaving a small gap between the heat break and PTFE tube. Clearly I must have cut the tube a bit short.
This was literally the only thing I could of, so I tried re-seating the tube all the way down inside the heatsink, reassembling everything, and I'm suddenly able to complete a test cube out of PLA.
Going to run a few more tests before I call this one resolved but this sort of fits what I've been seeing and is consistent with your advice about what could cause this.