MITM introduces significant delays, yet another single point of failure, adds significant risk of leakage. Current DPI are passive, so they have no such problems.
It's technical problem, that can be solved. Dedicated hardware optimized for MITM, probably.
It's already in use, in fact.
Who'd have thought banks would look for ways to save money?
Madness!
Could you elaborate a bit on what kind of passive DPI you can usefully do on an encrypted TLS session? Do you mean something like this? https://arxiv.org/pdf/1607.01639v1.pdf
Deliberately weakening encryption also adds significant risk of leakage, this time on a global scale.
There are absolutely "bump-in-the-wire" intrusion prevention systems with MITM capability.