While I agree that 100% test coverage is an ideal worth aspiring to, I think getting there from the current state is going to be a large amount of work that yields very little benefit at this point in time; I would say that there are more important things to spend that effort on.
What we're talking about is requiring new patches to cover the test code that they change, not an instant bar for 100% coverage of all test code.
Since this is being couched in terms of "effort" - it seems to me that this discussion alone is already more effort than just covering a few errant lines of test code here and there. :).
For now, let's just bite the bullet and require 100% patch coverage from here on out. If we hit a really nasty case where it really is a significant investment of effort, then maybe we can revisit this discussion and explore a better way to express this exception without losing information about test coverage completely.