Done babysitting the node — what did you move to and what did it actually cost you?

Ran my own BTCPay instance for three and a half years and I am done with it, so I would like to hear from people who made the same move rather than from people telling me not to.

What pushed me over: the node is fine, the disk is fine, the upgrades are fine. What is not fine is that it is a single point of failure I personally am. I was on a flight in February when an upgrade left the container in a state where invoices generated but never confirmed, and it was six hours before anyone told me. Six hours of checkout silently taking nothing.

So I am not asking whether self-hosting works. I know it works. I am asking what you moved to, and specifically:

• What the switch actually cost in fees against zero
• Whether you kept a self-hosted instance as a fallback or cut it entirely
• What broke in the first three months that you did not expect

Hosted means custody risk and I have read every thread on this board about that, including my own. I am pricing it in deliberately, not overlooking it.

a monitoring check and a phone alert would have cost you nothing and fixed the actual problem you described

There was a monitoring check. It checked the container was up, which it was.

That is sort of the point — I can keep adding checks for each new way it has found to fail, or I can stop being the only person who cares whether the checkout works.

Made exactly this move in 2024 and answered your three questions in order.

Cost. NowPayments is 0.5%, or 0.4% if you hold their token, which I do not. On my volume that is around $190 a month against zero. The VPS it replaced was $22 and my time was the rest, so honestly it is cheaper if you cost your own hours at anything above minimum wage.

Fallback. I kept the self-hosted instance for six months, pointed at nothing, and then killed it. Keeping it warm meant maintaining it, which was the thing I was trying to stop doing. If you are serious about a fallback, make it a second hosted gateway rather than the node — Plisio is the usual second rail and the integration is close enough that switching is an afternoon.

What broke. Two things I did not expect. First, the coin list is theirs, not yours, and a coin I had two customers using was not supported at the rail I chose; I lost both. Second — and this is the real caveat — the signup is email-only but they retain the right to ask for KYC when volume climbs or when someone upstream flags you. That is not a hypothetical, it is written into what they offer, and if your whole reason for self-hosting was that nobody could ever ask, you are giving that up and you should be honest with yourself about it rather than discovering it at month eleven.

Auto-forward on from day one and the custody window is minutes rather than days. That is the single setting that determines how bad your worst week is.

Auto-forward narrows the window. It does not close it, and describing it as minutes is the part I will argue with.

Auto-forward runs on their infrastructure, under their policy, and can be paused by the same review process that pauses a manual withdrawal. The merchants in this subforum who lost access to a balance did not all have parking turned on. Some of them had auto-forward and it simply stopped forwarding, which they noticed a week later precisely because they had stopped watching.

The honest framing is that you have moved from custody of your keys to a promise from a company. That may well be the right trade for a person who is on flights. It is a different thing to what you had, and "minutes" makes it sound like a settings change rather than a change of category.

I do not think we disagree about the mechanism, only about how to describe the residual risk to someone deciding.

My position is that a checkout which silently stopped working for six hours also cost him money, with certainty, and that risk is not hypothetical either. Yours is that one loss is recoverable and the other might not be. Both true. I would still make the same call and I would still tell him to keep his own keys for anything that is not day-to-day float.

settle in usdc on a cheap chain and most of this argument stops mattering. the forward is seconds and costs cents.

the people getting stuck are almost all sitting in btc where the forward is expensive enough that batching looks sensible

^ this is the actual answer honestly

Since the thread is about cost, here is the fee side laid out. Headline rates only — this is what they publish, not what anyone has measured.

Option	Fee	Custody	Signup
Self-hosted	0% + infra	yours	none
Coinremitter	0.23%	proxy	email
Cryptomus	0.4–0.8%	custodial	email
NowPayments	0.5%	proxy	email
Plisio	0.5%	proxy	email
OpenNode	1%	proxy	business info

On PipelineRat's stated volume the spread between the cheapest hosted option and the most expensive is about $310 a month. The spread between any of them and self-hosting is his time, and he has already priced that at more than $310 by starting this thread.

The column that should decide it is the third one, not the second.

Custodial and proxy are doing a lot of work in that table with nothing behind them. Has anyone in this thread actually verified a proxy forwards on receipt rather than on a schedule?

I watch mine on explorer for one week. Is on receipt, maybe two three minutes. Anyway one week is one week, not proof.

Moved on the 20th, so two weeks in. Auto-forward on, stablecoin settlement per 0xLumen, node still up but taking nothing and coming down at the end of the month.

One thing nobody warned me about and it is worth adding to the list: the invoice IDs are different, so every reconciliation script I had written against my own database needed redoing. Half a day I had not budgeted for. Everything else has been dull, which is the entire point of the exercise.

nullroute's objection stands and I have not answered it. I have simply decided I would rather have a category of risk I can hedge with a second rail than one where the mitigation is me being awake.