Here are some of my musings (not musigs:-) on this topic:
I think the most important milestone, at this point, would be to create an RFC for Multisig. Past experience suggests that implementations not closely following an RFC are much harder to review and leave questions about whether the choices made are really optimal for Grin.
I do not think we should pay anyone for creating an RFC. But once consensus is reached on an RFC and it gets accepted, then we can consider compensating its creators. And move forward with seeking ways to get the RFC reviewed by cryptographers and ultimately implemented.
Commenting on some of the above points:
There is no desire nor need for a scripting language in the consensus model. But some language for making wallets more
programmable could be useful.
I see no need for extending the MW protocol. Use of multisig should be invisible to full nodes. I expect the implementation will be mostly if not entirely within the wallet. And will likely require a new slatepack RFC as well.
Adaptor signatures are not strictly needed, but support for their use will be beneficial for applications like Atomic Swap and possible payment channels.
Note that Grin already uses multisig in the sense of sender and receiver both signing for txs. The distinguishing feature of this proposal is having outputs owned by multiple parties. Either n-of-n or more generally m-of-n. This necessitates multiple parties not only constructing a signature together (a mostly solved problem), but also constructing a range proof together.