The problem wasn't with speed, but with an envelope path where the two strokes had the same extremities (the red one for "SOUTHERN" below, so the script couldn't determine which was on top. of course I could tweak the path to avoid this but then I also use that envelope path to draw the "corners".
The new version just compares a pseudo-average height of the strokes (coordinates of the extremities plus topmost handle/tangent), which is simpler (otherwise 81 cases...)
Ah, but there is some even if not obvious, below, 3 paths, from top to bottom:
The envelope
Bendiness 100% (red)
Bendiness 0%
As you can see, bendiness 100% added tangents (at 100%, the tangents from from both anchors are exactly on the middle of the segment and so overlap). To quote the doc: "Note however that this curve cannot be exactly the same
as the one you would get by interpolating intermediate points (even if it will be close)."