Been thinking about using NDepend to flag certain conditions post check-in as a help, but I don’t think it completely replaces the need for a good code review.

Ideally we’d do pair programming but I’ve never been able to fully sell that principle into the system.

Thanks for the suggestions.

Yeah, unfortunately we are always in the situation of delivering what the customer wants or delivering the best engineered solution. They almost never meet.

In terms of your problem of discovering bad code just before release, the first thing that comes to mind is perhaps to have your code review earlier in the cycle, i.e. just after check-in or just before.

The second step would be to introduce some check-in policies which the developer(s) must adhere to before checking in the code. Some tooling can also be introduced to automate some of the checking for you.

Hope it helps and good luck.

Cheers