Renata Fitch

Easiest fix: put the classic menu back with the registry key. Menu becomes instant, problem gone. Been running mine that way for ages.

Fair. Just tested on the work laptop and the classic menu comes up about as slowly as "Show more options" did, so I was measuring the click I'd removed, not the load. Withdrawn.

The other version of this mistake: tracert resolved a name to a different address than the one you pinged, so you compared two different destinations. -d and an explicit IP on both commands, always.

Set timeout: 60000 in the config and move on. CI is slower than your laptop. That is the whole story.

This is the strongest argument for a subscription vendor that I know of. Not because acknowledging is hard - it is one call - but because it is exactly the sort of rule you only find out about by losing money once, and there are maybe fifteen of those across the two stores.

Do not cancel with proration. Stripe will happily generate a credit note and park the balance on the customer, which does absolutely nothing for someone who is never coming back. Cancel, then issue an explicit refund against the charge. Credit on a dead account is just a way of not giving people their money.

Four to eight hours is completely normal for a first build. The videos are people who have done thirty of them, with the parts already sorted, editing out the boring parts.

Where your time actually went, roughly: switch soldering is 136 joints and at a genuinely careful pace that is an hour and a half. Stabilisers, if you tuned them, is another hour on a first attempt. Testing and inspecting is where beginners spend the rest, and that is time well spent.

An order that saves an hour next time. Test the PCB with tweezers before soldering anything. Do the stabilisers completely, and test them by hand, before any switch goes in, because reaching a stab under installed switches is miserable. Put in the four corner switches, snap the plate on, check it is square, then solder those four, then the rest. Flash the firmware before you install every keycap.

Second build will be two hours. Third will be ninety minutes and you will stop counting.

$340 a month is $4k a year for basically nothing. Servers do not care. Keep it forever.

Generalise it so you stop thinking about it per-handler: every job carries an idempotency key from the producer, and you have a processed_jobs(key primary key, completed_at) table. First statement in the handler inserts the key, in the same transaction as the work. Conflict means you are a duplicate, return.

Boring, portable across every queue you will ever use, and about fifteen lines of shared code.

Unpopular: do not mock the database. Use an actual Postgres - a container, or one of the in-process builds - with your real schema and a transaction or a fresh schema per test.

A mocked query builder means your tests pass while your SQL is wrong, and wrong SQL is the only thing database code can actually get wrong. You are mocking out the part under test to make the part you already trust easier to arrange.

Did the tape, recorded before and after. Higher pitched, slightly poppier, hollowness untouched. Which is exactly what you said would happen.

That also destroys the jobs that were perfectly fine, every repeatable and delayed job definition, and anything else sharing that instance. getJobs(['active']), look at them, move the genuinely stuck ones back to waiting. Flushing a queue backend is the kind of thing you do exactly once and then never again.

Can confirm from doing exactly this. Look for a relay with browned board underneath it and a slight burnt smell. Desolder, replace with the same rating from any electronics supplier, and the fault goes away. Took me an hour and about four dollars including postage.

Rattle that survives lubing is nearly always slop, not friction, and one sided rattle narrows it a lot.

Check in this order. Is the right hand stab housing fully clipped into the PCB and the screw actually tight. A stab that is 90 percent seated on one side produces exactly this, a silent left and a rattly right. Second, are the stem feet clipped. Unclipped feet let the stem rock in the housing and no amount of grease stops that. Third, the wire. Lay it on a flat surface and look along it; if it rocks at all it is not flat, and the end that sits proud is the end that rattles.

The holee mod is what fixes wire slop properly: a small piece of thin fabric or bandage material wedged into the stem insert where the wire sits, so there is no free play at all. It is fiddly for ten minutes and it is the difference between quiet and silent.

First submissions do seem to get a rougher pass, and enough people have a strange first round that it gets treated as a rite of passage. It is not one, but do budget one round of back and forth into your launch date rather than assuming a clean run.

Mild disagreement. API seeding is three or four times slower and on a 40 spec suite that is minutes. Direct SQL for the boring background furniture, API for whatever the test is actually about. Best of both and the invalid-state risk stays small.

For next time: hotswap sockets are held on by solder pads about the size of a grain of rice and they hate lateral force. Always check the pins are dead straight against something flat, always press straight down, and always support the back of the PCB with your other hand or lay it on a hard surface. Five pin switches with the plastic legs need to line up perfectly or they lever the socket sideways as they go in.

Good chance you can see it under a phone camera zoomed in. A lifted socket sits a fraction of a millimetre proud and the solder fillet looks cracked and grey rather than shiny.

Before you bin the next one, open it and put a multimeter on the cells. A pack that dies after four screws is almost always one cell group that has dropped out of balance, not five bad ones. If you see four groups at 3.6 volts and one at 2.1, that is the whole fault.

Rebuilding with new cells is doable but you need a spot welder, or a friend with one. Soldering directly to cell tops cooks them and it is genuinely a fire risk, so do not be tempted. If you are not set up for it, at least you will know whether the pack was salvageable before you spend on a new one.