On commitment tx fees :
instead of having the participants of a channel agree on a fixed predetermined fee between 2*e and 10*e, where e is estimation to enter next block;
have you considered : participants of a channel agree on multiple fees for each new channel state ? Each time Alice and Bob agree on the new channel state [a;b] they would sign multiple commitment tx [a-0.5*e;b-0.5*e], [a-e;b-e], [a-1.5*e;b-1.5*e], [a-2*e;b-2*e], … which would enable RBF commitment tx (avoid over-paying and avoid having stuck tx).
By the way, thanks for this great post, enjoyed reading it.