Fees are down to 1sat/b. My transaction finally confirmed.

My lightning node had to force close a channel and the transaction finally confirmed. On March 10, my channel partner lost his channel state and so my node broadcast a force closing transaction:

2021-03-10 16:17:16.494 [ERR] LNWL: ChannelPoint(7423dc29...:0): sync failed: remote believes our tail height is 0, while we have 1200! 2021-03-10 16:17:16.570 [INF] CNCT: Attempting to force close ChannelPoint(7423dc29...:0) 2021-03-10 16:17:17.255 [INF] CNCT: ChannelArbitrator(7423dc29...:0): force closing chan 2021-03-10 16:17:21.104 [INF] CNCT: Broadcasting force close transaction 70189d1..., ChannelPoint(7423dc29...:0): (*wire.MsgTx)(0xc00cfa2fc0)({ 

Now the transaction confirmed:

2021-05-12 02:05:05.402 [INF] NTFN: Dispatching confirmed spend notification for outpoint=7423dc29...:0, at current 683188: 70189d1...[0] spending 7423dc29...:0 at 683188 2021-05-12 02:05:05.976 [INF] CNCT: Local unilateral close of ChannelPoint(7423dc29...:0) detected 2021-05-12 02:05:06.060 [INF] CNCT: ChannelArbitrator(7423dc29...:0): local on-chain channel close 2021-05-12 02:05:08.535 [INF] CNCT: ChannelArbitrator(7423dc29...:0): trigger localCloseTrigger moving from state StateCommitmentBroadcasted to StateContractClosed 2021-05-12 02:05:08.535 [WRN] CHFT: could not remove channel: channel: 7423dc29...:0 not present 2021-05-12 02:05:10.133 [INF] CNCT: ChannelArbitrator(7423dc29...:0): still awaiting contract resolution 

The output is still time-locked for a few blocks (the confirmation time counts, not the broadcasting time). Then my node will finally be able to claim the funds.

Looks like lightning server authors still have problems to get the fees right. There was nothing that I could do to accelerate the transaction (CPFP doesn't work as the output is time locked). Luckily there wasn't an HTLC output involved, where the confirmation is time critical.

I wouldn't fault the lightning server authors, alone. True, in this case the node should have rejected the last channel update in January as the fee of 1 sat/byte was clearly to small. But it is really hard to get the things right. If you close channels preemptively – because you think there could be a congestion and the other node hasn't updated the fee – that can also be harmful. If you read the lightning whitepaper it makes it clear that some security assumptions rely on predictable confirmation time on the base layer.

Anchor outputs will hopefully fix this at the cost of making force closing transactions a bit more expensive for the extra outputs. But at least you (or rather the software authors) can better fine-tune the fees without relying on cooperation from the other party. Though often, I find the automatic fee estimates a bit too high and I have spent $4.50 fee in the past to recover 80 cent from a $5.30 channel output (which means I basically spent everything on fee, unless I can move the 80 cent at 1 sat/byte).