I think there's a difference between a community site for code reviews, and a commercially viable hosted code review service.
For the former, I think that's a great idea. Last year I tossed around the idea of hosting Drupal code review meetups, where code to review would be submitted ahead of time with notes, then walked through live and the attendees (who would presumably have made notes prior to arriving) would comment on it. People who had code reviews in previous meetings would return to present their changes, comparing and contrasting.
The tricky situation with code reviews by people who are unfamiliar with your project is there's a lot of context that they lose out on. A jack-of-all-trades programmer is only going to be able to discuss broad technical constructs and algorithms. An expert PHP language programmer is going to be able to contribute more specifically. But only an expert Drupal developer is going to have enough context to really significantly improve your code, second only to someone familiar with your codebase's internals. You just have to know what you want to get out of the reviews.
As for a commercially-viable reviewing service, I'd wonder what sort of developers you'd have doing your reviews. I ran a developer relations team for a few years and we needed to hire serious, top-tier developers (and pay them top tier rates) to really get the kind of in-depth understanding and support equivalent to the original engineers. We spend easily a man-day across 2-3 developers for each code review we do, sometimes longer. That's going to be expensive.
In addition, I'd be concerned as to why your customers were using your service. If it's a form of "sign-off" or "best practices" review, an impartial third party before a delivery from a consultant to a client, I'd be worried for the reputation of your service. I've hired developers and put their code through code review and we all said it looks pretty good but it wouldn't actually run or perform as advertised. If a shady consultant got your service to say they were great, but it turned out their client was scammed, the client may come calling for you.
Finally, we use Smart Bear Code Collaborator. It's not bad, but it's not perfect. Maybe an MVP is simply cleaning up something like Rietveld and competing with it or Kiln as a hosted service, first.
Good luck!