I may not be the best person to answer this, but in the original GOOSE-1, bring-your-own-validator (BYOV) modules were already mentioned as a promising path. The idea is to make modules a lot more granular than they are today. At the limit, you may have one module per BYOV staker, and this staker can decide who that module’s node operator is.
Having a non-Lido node operator raises new challenges around MEV stealing, performance, risk management, and more. Still, these challenges are not too dissimilar from those Lido contributors face in permissionless CSM modules. It is a theory that such a design can be further generalized to cover verticals like restaking as well.