It is new to me, but raises a concern that they could only review for general engineering quality, not whether or not a PR is appropriate to a codebase. How would they know if the same problem has already been solved elsewhere, that no wheels are being re-invented, or that a PR is stepping on the toes of something else the team is working on.
Because that is the value I've found in code reviews - not generic "is this code elegant?", but "does this code play well with what everyone else is doing?"
I guess you could first have an internal architect or similar vet the PR before handing the PR to this service for the "technical details".
As you said, the big value in reviews are the points you mentioned. Correctness/technical quality certainly has value, but at $700/dev/month the reviews better be really good. Especially since doing it internally has value as well (knowledge sharing in particular).
Because that is the value I've found in code reviews - not generic "is this code elegant?", but "does this code play well with what everyone else is doing?"