Heyo:
It's not the converter, as your problems on the top row give away what is happening.
Unfortunately this is particular to the XT era boards as previous beam-spring and subsequent F used separate controllers. The XT's controller is on the same pcb as the main matrix, and while it is digital and not particularly vulnerable to EMI - it can emit it, so you'll see it fully shielded with copper foil for EMS applications, say, for use near an old EKG in a hospital. (One of the hospitals I had a scan at still used the old faraday cage room that was needed in the early days.) Such shielding is no longer needed, but it was around 1980.
Point here was that just by taking a look at how those are shielded in some of the posts here will give you ideas how to wrap up your beasty, without worrying about details... or so I thought.
IMO the reason this is particular to the top row on the XT is that the top sense line is exposed to a ground plane by running parallel to it for 14" - other F don't treat ground planes as magical or safe.
I don't use a pro-micro, you'll need to look and see if that isn't connected to the logical ground already - if were, I might have suggested you separate them - but it doesn't matter - the XT ties them together. This takes place at the screw on the controller, which is visible when you open up your XT.
Rather than trying to separate the logical and case gnd all along the chain, I'd just release that one panel from the local gnd - since I expect it the interference is coming along the gnd, via the shielding. The panel will be exposed - but only for 15cm (6"), so we're good.
So - the screw head will partly disconnect the case (meaning the steel plate on the bottom of the keyboard unit, and the plate-with-holes overtop, not the plastic top-shell and bottom steel base full of crumbs and spiders.)
You could fully disconnect the case from everything else, by placing something under the pad under the screw hole - between the pcb and the backplate of the case. This might help, if the interference is being picked up by the case mostly - but I expect it is coming along the USB shield and ground lines.
I am confident that the sketchy row will behave as well or as poorly as the others if you free that problem section of ground plane from the rest of the system. To do this, the screw needs to be removed, though it's up to you whether to leave the bottom pad connected, as it connects to the other side of the pcb... and, the two traces from the problem plane directly above the screw, connecting it to the logical (and thus USB) gnd need to be cut, as well as the two vias connecting the section of plane to gnd on the other side of the pcb. The two vias are to the right, about a half inch apart, just before the 8 sense lines drop down under the top of the case. This is all on the screw-head side of the case, i.e. the top side of it, which is the solder-side of the pcb, not the component-side. Oops - and dont forget to get the little trace running up from the second via beside the 8 lines - it runs up to pin 20 on the IC, needs to be cut as well.
Great thing about this, is you can do it without opening the irritating part of the case and getting keys and springs everywhere.
I'll add a pic in a bit, I expect - but my pcb at hand has been slightly modded - be warned.
Hope to be at least entertaining,
dfj