Why does my N4M only print clean after the first layer?

Since your straight walls are printing clean, I'd stop worrying about the leadscrews and gantry squaring. A bent or wobbly screw marks every part of the print at a regular height interval, flat faces included, so that's not matching your symptoms. Same logic for the gantry. Here's a quick way to confirm: shine a light at a low angle across one of those curved sections and look at the ridge spacing. If the ridges repeat at the same distance apart all the way up the part, that's your Z screw. If they cluster wherever the nozzle changes direction and fade out along straight runs, it's vibration from speed and cornering, which is what I'd bet on given how curves behave versus flats.

Curved walls are basially a constant string of direction changes, so the head never gets a chance to settle, while a flat wall only gets hit at the corners. That's why slowing your outer wall helped a little but didn't kill it. Speed isn't the only knob. Look for the acceleration or cornering setting in your slicer profile and bring that down on the outer wall too. The stock profile on these runs pretty hot, and on round stuff the head is whipping through turns constantly. While you're at it, pluck both belts like a guitar string. They should have some firmness to them, not feel floppy. If one side is noticeably looser, snug it up, but I'd guess this turns out to be a slicer change more than a mechanical fix since the bottom layers are dialed and squares are square.

One more thought: coats of that 'smooth by default' look people get on tall cylinders ussually come from going slower on the outer wall AND easing the cornering, together. Mine took a couple of test cylinders to nail down, but once I dropped both values the ridge pattern disappeared entirely. Print the same tall round model twice, once with each change isolated, and you'll know which one is doing the work.

Had the same mystery on mine last year and it turned out to be dumb simple. The spool was dragging on the side of the holder and stalling just enough to mess with the feed on curves, where the head keeps changing speed. Print looked fine on straight runs, rough on round stuff. Swapped to a smoother feeding setup and it mostly went away. Also check the little wheels on the carriage for flat spots, mine had one from running tight and it showed up as a faint bump every rotation. Cheap fix either way, worth a look before you go deeper into the slicer settings.

Loose pulley did this to mine. The grub screw on the X motor pulley had worked itself loose just enough that every direction change gave a tiny lag, and curves are nothing but direction changes. Straight runs hid it completely. Snugged it back down with a drop of thread locker and the wobble was gone. Takes two minutes to check so worth doing before you go deep into slicer settings.

Chiming in with a diffrent angle since the pulley and spool stuff is covered. Mine had the same curve-only roughness for weeks and it turned out the printer was sitting on a flexy folding table. Any wobble in the surface gets amplified on curves where the head is always changing direction. Moving mine onto a solid bench and putting a cheap foam exercise mat under the feet was the thing that finally killed the ringing for me. If your printer is on anything that flexes at all, try pressing on the desk corner while it's printing a curve and watch the surface.

Update: turned out to be the spool thing. Checked mine and it was rubbing against the side of the holder, dragging just enough to mess with the feed on curves. Put a smoother washer setup on it and the ridges are gone. Feel a little dumb, since I'd been tightening gantry bolts all weekend for nothing. Straight walls and curves both clean now.