Ned

Your pace is fine, but check your technique is not the reason for it. Tip at 320 to 340 degrees, tin the tip, touch the pad and the pin together for about a second, feed solder for another second, remove solder then remove iron. A good joint looks like a small shiny cone. If you are sitting on each pin for five seconds waiting for something to happen, your iron is too cold or the tip is oxidised, and that alone can double the time.

A travel router will not fix hotel congestion. What it fixes is the captive portal, letting all your devices on with one login, keeping your traffic on your own subnet, and giving you failover if it takes a USB modem or a second SIM. Those are real, but they are not your symptom.

Your symptom, freezing every ten minutes with fine speed tests, is packet loss or a roaming event, not bandwidth. Speed tests average over ten seconds and hide both. Diagnose it properly before you spend anything: run a continuous ping to a reliable address and leave it running through a call. If you see loss spikes at the moment the call drops, it is the network. If you see the ping stall entirely for two seconds and then resume, your laptop is roaming between access points, and the fix is to lock onto one band or one access point rather than letting it wander.

In hotels I get most of the way there by connecting to the 5GHz network only, sitting nearer the corridor access point, and if the room has an ethernet socket, using it. Wired hotel ethernet is often untouched and excellent because nobody uses it any more.

Recording it is the tip. Every mod sounds better immediately after you do it, because you just spent an hour on it.

Before spending anything, swap keycaps if you have a second set. Going from thin ABS to thick PBT or vice versa changes the sound more than any foam decision, and you already own the test.

On grease choice, use a thin dielectric in the housings where the stem slides, and something much thicker on the wire ends only, the consistency people use for stabiliser wires rather than switch sliders. Thin grease on the wire ends migrates away within a week or two and the rattle comes back, which is why it seems to work at the bench and fail three days later.

Clipping the feet is the step everyone skips because it feels destructive. Nail clippers, take the two little legs off flush, it changes the sound immediately.

A whole row failing at once is a matrix fault, not twelve simultaneous switch faults. The switches are almost certainly fine.

In a keyboard matrix, every key in a row shares one trace and usually a set of diodes. If one point in that line breaks, everything downstream of it goes quiet. Your bent pin story fits: the extra resistance you felt was the pin catching the socket, and the socket has almost certainly lifted its pad off the PCB, breaking the trace for the whole row.

To confirm, take all the bottom row switches out, open a text editor, and short each socket by bridging its two contacts with tweezers. If nothing types anywhere in the row, it is the board. If one particular socket also looks slightly raised or tilted compared with its neighbours, that is your culprit.

Fix is usually a reflow of that socket, and if the pad has actually torn off, a short bodge wire from the socket leg to the next point on the trace. Neither is hard, but you need an iron.

Worth adding that these settings interact, so change one at a time and type normally for a day between changes. Changing three at once and judging by feel gets you nowhere.

Instrument it rather than guessing. Turn on console output and watch the key events as you type the failing sequence; you will see the exact interval between press and the next keydown, and then you know what to set the term to instead of picking numbers. Takes ten minutes to set up and saves a week of vibes based tuning.

That explains a lot. They are on the charger permanently because I got sick of finding them flat, so I optimised for exactly the wrong thing.

Which is also why the first launch of a rarely used app after an update can still be slow. That one had not earned its compilation yet.

Same family of idea as a desktop rebuilding a search index after a big update, or a game shader compilation screen before you play. All of them are one time work done up front so the thing you actually do afterwards is fast.