I think such a plan is something the witnesses should come up with. In my pledge I do just that:
Always have one node on the old version, should problems arise, and be prepared to roll back changes.
So I will have a plan to roll back changes and will communicate that with other witnesses. Best case we never need to roll back ;)
I think those are two different things. You have a plan for how to roll back, which is great and necessary. But we need an explicit plan for making the decision to roll back as well.
If 99% of users can't interact and 80% of the voting power disappeared, should we roll back? I think if we asked that question last week, virtually everyone would have said yes. Instead it was only asked after it happened, so we didn't get a rollback.
I think such a decision would happen organically if all witnesses have a plan. Just as any other hard fork decision happens.