Context Management tests - follow-up, can we use the old model?

[afarntrog]

As a follow-up, can we use the old model?

This is just more expensive for no purpose AFAICT

Originally posted by @zastrowm in #2193 (comment)