Bastien Teinturier

> I don't see how they're related or even touch similar parts of code. It does require more code, at least in our implementation, and it requires expanding the test...

> Eclair and LND done yeah? Yes, it has shipped in lnd and eclair a while ago. As for cln, they need to do some work on the fee-bumping utilities,...

> I think, especially for mobile/0conf/SCID aliases, we'll probably always need to support no-anchors. Right, it could make sense to add support for other commitment types than `anchor_outputs_zero_fee_htlc_txs`, I'll give...

IIRC there was a lack of logs on the LDK side to properly figure out what happened, right? If anyone cares enough, it should be easy to run the same...

Great, thanks for testing it @sstone! Do you have suggestions on how I could unit test that, or should we merge this without tests?

Don't you think [this explanation](https://github.com/ACINQ/eclair/issues/2388#issuecomment-1220467615) correctly describes the issue in #2388? That's what lead me to make the change in this PR.

> It makes sense but then how did we identify that such channels were actually not stale, and why did they get pruned so quickly ? The way I understand...

@sstone @pm47 I've updated our logic to store more information in the pruned DB to ensure we don't constantly restore channels that are actually half-dead just to prune them again...

> This works, but I think we can go a step further and remove the pruned_channels table. That's a great idea, I started implementing this and it's looking good. I'll...

Superseded by #2456