If there are failure scenarios then I don't think it will work very well, nor do I think many people will be willing to even give it a try.
There are failure scenarios in everything. Not having segwit for LN introduces a few more possibilities for failure (if coded improperly) than without segwit. Not having segwit requires more complexity and thus a higher possibility of a bug.
I have thought about this, and I cannot think of a way that LN can work without malleability being removed from Bitcoin, and without additional forks of any kind, and meeting the criteria of the ability to have instant, trustless, two-way transfers. If you want to explain how LN might work, then be my guest.
From a technical perspective, in a vacuum, LN should work once malleability is removed, assuming a client can be programmed in a way that it cannot be tricked into signing transactions "out of order". From an economical perceptive, I am not so sure for a couple of reasons.