The IPv6 flowlabel field can be used by NICs for RSS. In SRv6 encapsulation, this is crucial as most NICs likely will only be able to use the outer IPv6 header to pick the queue, so without a flowlabel, all packets with the same outer IP addresses will end up on the same queue.

From @YutaroHayakawa:

The H.Encaps section of RFC8986 says it should follow the way specified in RFC6437, but its 🤯
I think we can follow the way of kernel implementation.
torvalds/linux@b5facfd#diff-4da369b586ba9233e8a230875d06c5daa731175a8c483e3fcc0691aab9cd9817R103-R105

Following the kernel implementation is probably a better idea indeed. We could more specifically follow the seg6_flowlabel=1 implementation. Not sure if we have an easy way to retrieve the skb hash though (i.e., not sure skb->hash is always set at the time of encap.).