Tl;dr: It's *much* better, now.

What'd I do?

Ah, yes... first I tried three parallel diagonal lines filling most of the page. The thought was to see if the roller itself was somehow binding; then the lines would show errors in the same locations, right?

The first line was horrible... wobbling all over the place. But, when it finished I was greeted with a loud grinding sound of doom... which... turned out to be the sound of stepping backwards up the page. Apparently, confirming an earlier observation that when performing a backward motion it actually oversteps, then moves forward again. I guess the idea is to compensate for backlash in the gears. So, a 1/120in step /up/ the page appears [later verified by 'scoping] to be like 12 steps up, 11 steps down.

And... my moveTo() function does-so not by telling the printer where to move, but by telling it to move 1/120th of an inch, however many times are necessary. So, heading back 8inches up the page... well... it was slow and loud.

But, the /second/ diagonal line, printed in exactly the same manner as the first, dot-by-dot, step-by-step, in the same direction, just half an inch to the right... well... it turned out perfect. And so did the next.

...

here, the theory goes that there may've been debris in the gears... which seemed confirmed, later, as well. And that all that rattling shook it all loose!

OK...

TONS of other lookings-into, including 'scoping...

E.G. they used what seem to me *tiny* traces to carry half an amp 10+ inches... and small vias... and... yes... [probably most importantly] one of those vias happened to be in the electrolyte-flood-zone... I measured nearly a volt drop between the power source and the motor. At 14V, that may not seem like much, but *power* is what does work, and P=V^2/R... V-squared... 13^2 vs 14^2 is not insignificant!

Rather'n just fix the via, I ran heavy wire from the power supply straight to the motor-driver/connector.

I also looked into the service manual a bit... the same motor is *also* responsible for driving an optional sheet-feeder... so, this thing *should* be able to handle quite a bit more loading...

Also, finally got a closer look at the actual gearing... yeah, white debris, presumably from the white gear, on the brown gear attached to the paper roller. Prb not the best sign; maybe 1000's of 1/120th inch start-stop motions in both directions is a bit much to ask of an inkjet. Also, interestingly, the motor's brass gear seems worn more than the white plastic one it drives[?!]

OTOH, another discovery: the white gear actually slides on its shaft a bit [maybe 2mm]... this is enough that in one extreme maybe 1/4 of its thin-enough tooth-thickness hangs off the driving gear, whereas in the other extreme it's got plenty of wiggle-room to spare while still being fully-meshed.

I didn't do anything about that...

But I ran the ol' star experiment again... and; well this won't make sense unless I mention that I ran a few more test-prints after the highly-successful line-test, but before attempting repairs... and... yeah, it started up missing steps again.

So, /after/ the new wiring and fiddling with [but not fixing] the gear-sliding... I ran the star experiment again... twice... and they both were *perfect*. Which I'm not sure I've ever seen.

So, then I thought about that gear sliding, and tilted the thing on its side... [gear sliding slightly off the motor].... and a mess of missed-steps, again. [None, surprisingly, by the carriage, which now has to lift against gravity]. Trying the other side, though, wasn't really feasible; part of the clutch-system relies on a spring which is too weak to counteract the weight. I managed to give it a little push *once* at the right time and it managed printing most of one line before aborting... that line, too, indicated missed steps on the up/down axis [but again, not on the left-right axis now fighting gravity!]. Now, I suppose, it may just not've slid that gear back into place...

Then I did two more tests with the printer in its normal orientation... and... perfect, both.

Heh.

Falling asleep, I'm sure I've forgotten a lot