Summary of discussion with a routing node operator:
-
It may not be necessary to signal HA per channel. On the node level in the node announcement may also work. Note: without code changes, an alias suffix
(HA)may be an option too? -
It may not be necessary to signal a guaranteed amount that the node should always be able to route. Can also leave it to the node operator to make sure liquidity is enough for the traffic that they see.
-
Part of the strategy of a HA node could be to stop signalling depleted channels.
-
Nodes apparently start to experiment with returning errors different than “temp channel failure” to try to avoid penalization. Note: they should probably just return a corrupt failure message which currently can’t be attributed, to evade penalization even more.
-
Node sometimes totally shut off routing for a while. Can be when chain fees are high and they don’t want to risk force closes.
-
HA nodes could also be selected preferentially for blip39 blinded paths, to increase the likelihood of success
-
Having proper inbound liquidity could be part of the requirement to be a HA node. So this extends the responsibility of the node.
-
Some nodes (especially tor) are really slow and it is currently not possible to pin point them.