The lightning network as is is working without SIGHASH_NOINPUT. Thus in its current form it is not needed. 

However Christian Decker came up with the eltoo payment channel construction. In a nutshell this would get rid of the penalty based construction we use right now and reduce a lot of overhead for maintaining channel state as the state becomes symmetrical for all participants. 

In particular multiparty channels and channel factories would become much easier. 

The eltoo paper however really depends on SIGHASH_NOINPUT as the channel state is encoded as a "linked" list of update transactions which in case of protocol breach are skipped from the breaching point to the current channel state which requires SIGHASH_NOINPUT. 

It is my understanding that essentially all lightning network developers agree that lightning network channels should eventually be constructed via eltoo. 

You can read a [summary of eltoo on my blog][1] or you can [read the original publication][2] to learn more.

[1]:https://www.rene-pickhardt.de/index.html%3Fp=2116.html
[2]:https://blockstream.com/eltoo.pdf