ISTM that the tests are *passing* spuriously -- they need to be less trusting that a single valid snapshot implies stability -- and that the real problem is indeed that the dummy provider wasn't updated to handle multi-model properly.
I would *love* for us not to have to touch the dummy provider; does anyone see an alternative?
ISTM that the tests are *passing* spuriously -- they need to be less trusting that a single valid snapshot implies stability -- and that the real problem is indeed that the dummy provider wasn't updated to handle multi-model properly.
I would *love* for us not to have to touch the dummy provider; does anyone see an alternative?