Owen

we did the big bang on 200 files, took a weekend, merged monday morning before anyone was in. worked fine

A sweat cover over the top tube and stem plus a towel over the bars costs very little and catches most of it. They look faintly ridiculous and they work. Combine with a fan and wiping down, and the whole problem essentially goes away without buying a second bike.

Seed through your own API rather than raw SQL. Insert rows directly and you will eventually create a state your application cannot produce, then spend a day chasing a bug that only exists in the fixture.

It's not soft at all. Reducing your required monthly outgoings raises the floor under every other decision you make for the next four years.

The framework I'd use is: what does the debt do to your flexibility, not just to your arithmetic.

A £300/month payment is a fixed claim on your income for four years. If you lose your job, the investment portfolio is optional and the loan isn't. Paying it off is buying certainty and reduced monthly obligations, which is a real product even though it doesn't show up in a spreadsheet return.

That said, 4.1% is genuinely low. If it were 9% this wouldn't be a question. At 4.1% with a full emergency fund, this is close enough to a coin flip that I'd let the tiebreaker be how much you dislike the payment: and you've already said you hate it. That's a legitimate input.

Insurance is where I would be most careful, and I would not take a workaround from a forum on this one. Vehicle insurers price on where the vehicle is normally kept overnight, and an address that is not really yours is the sort of thing that surfaces at claim time rather than at quote time. I ended up phoning a broker, explaining the actual situation, and paying somewhat more for a policy written for it. It cost me about a fifth more than my previous premium and I stopped worrying, which was worth more than the difference.

Asking for the refusal in writing changed the answer for me on the spot, without any confrontation. I suspect it is often a front desk assumption rather than a practice policy.

Then the broader one is arguably doing something the other isn't, even if the returns track closely. Their behaviour diverges most in exactly the periods you'd care about.

Worth checking they're genuinely near-identical rather than just having the same top ten. Two funds can share their largest holdings and still differ meaningfully in the tail: one might include mid caps, or exclude a whole region. Compare the number of holdings and the regional breakdown, not just the top ten table.

Go where the problem is already being described out loud. For bookkeeping cleanups that's accountants, specifically small practices who take on a new client, open the books, find a mess, and don't want to spend chargeable hours untangling it. Ring five local practices, say exactly that in one sentence, offer to do one at a reduced rate so they can see the work.

One accountant who trusts you is worth about thirty local Facebook posts. And the referral is warm in a way that a group post never is, because it comes with implied vetting.

They're the opposite. Most small practices actively don't want the low-margin tidying work, they want a clean set of books to file from. You're removing their least profitable hours.

Doing types slightly wrong for six months is how literally everyone learns this language. The fact that the cast felt bad enough to ask about means your instincts are already ahead of your knowledge, which is the right way round.

Postgres queues fall over somewhere around 100 jobs a second. It is a known ceiling and you will hit it eventually and regret the whole thing.

One accountant, one hour, cheaper than you fear. Say exactly what you said here. They deal with this weekly.