Is segregated witness necessary to implement the Lightning Network?

Lightning Network (LN) uses payment channels with Hashed TimeLock Contracts. Both of those things are currently usable on Bitcoin mainnet without segregated witness.

However, without segregated witness or another signature malleability fix, LN channels have to deal with situations where transactions get mutated (“malleated”), which makes them get stuck at various steps. Preventing them from getting stuck permanently requires either introducing trust (which we don’t want to do) or setting some annoying timeouts that limit the efficiency of channels and downgrades the user experience.

Some other relevant benefits of segregated witness in the context of Lightning are:

Channel monitoring: Users of Lightning typically have to watch their channels in order to avoid letting their counterparty broadcast a previous state, effectively defrauding them. To prevent against this users require that the txid of the transactions involved be known in advance and preferably immutable so that, once handed over to monitoring peers, additional user intervention is not necessary and penalty transactions can be enforced.

More block space: Meaning more channels can be opened or closed in any particular block.

Also, fixing malleability for upgraded software makes designing LN-compatible (and any Bitcoin-compatible) wallets much easier in general.