>He suggests a particular method for doing code reviews.  For a very
>small company like mine, I've had to adapt the methods.  But I'm
>getting good results finding latent bugs that would not have been
>caught in testing.

Do you proceed to do a specific test to verify you actually have a bug when identifying
one of these situations? If not it seems to me that you may be breaking something that you
only think is broke, but may not be.