The cheaper operator changes with the direction

Same pair, same amount, same minute: one of them is half the price of the other. Not a basis point — a factor of two, on a swap where neither of them holds the money and the output is identical.

The order changes. Going in, Coinos wins by a factor of two. Coming out to Liquid, it loses — 0.547% against 0.447%. Coming out to Bitcoin it wins again, but by 11% rather than 100%.

Three comparisons between the same two operators, and the answer is different in all three. A page that named one of them “the cheap one” would be wrong two times out of three.

Read Coinos’ percentage down the two tables: 0.1% to enter, 0.5% to leave. The same operator, the same chain, the same size. SATS Routing charges 0.5% and 0.6% for the two directions — nearly flat.

So Coinos’ entire advantage is on the way in, and it is a pricing decision rather than a property of Lightning. Somebody who measured once on the way out and concluded the two operators were within 11% of each other would have got the answer right for that direction and missed a 2:1 gap in the other.

The Lightning page already publishes that the cheapest operator inverts with size, because a percentage fee and a flat chain fee cross over. Direction is a second axis on the same table, and it is not the same axis: this one is a price list, not arithmetic.

Boltz          does not resolve ZEUS Swaps     answered, with no pairs Middle Way     HTTP 401  {"error":"unauthorized"}

These are not one outage repeated. A domain that does not resolve, an endpoint that answers normally with an empty list, and one that requires a credential it does not advertise are three different failures, and only the first is the one people mean when they say the Lightning swap market shrank.

Boltz is the one worth naming. Its backend is what several other operators run, which is why so much of this market looks like one implementation behind several brands. It is also, this morning, the one whose host does not resolve — and an empty pair list from ZEUS reads on a dashboard exactly like a working operator with nothing to offer.

Nobody publishes which operator served a given Lightning swap. Wallets route through one and do not say so; the fee appears as a fee. The figures above exist because this site asks all five on every load and shows the answers, and they are a snapshot — the operators change them, and nothing here keeps a history of it yet.

That history is the obvious next thing and it costs nothing: the answers are already being fetched. Uptime, fee drift and refusal history per operator, the way the swap board already does for providers, would be the only public record of post-Boltz Lightning liveness. The same argument, and the same method, as the swap price dataset — keep the losing answers, print the denominators, publish it free.

Measured 8 November 2026 at 100,000 sats. Two operators, seven quotes, three non-answers. Re-run it before quoting it; the point of saying when is that it moves.