To gauge better the fault-tolerance of this construction, I think you should have an estimation of the on-chain weight units costs for a range of less than perfect interactivity situations (e.g m-of-m, m-of-m - 1, m-of-m - 2). Also any assisted exit sounds a double-edged sword mechanism as it could be leveraged to kick out a user, at the worst time e.g when the on-chain fees are full.
There is no guarantee that the LSP respects its own liquidity policy on that regard.
In order for
A
to exit, it must first have theA & L
channel funding outpoint confirmed, then finally exit theA & L
channel using standard Poon-Dryja unilateral exit. Thus,A
must publish the path to its channel below:
Another question - How do you prevent A
to unilaterally downgrade the construction fault-tolerance for all the subset of users intersecting with its path to its channel ? I.e A to D being partitioned from E to H.