Usability testing is a part of the agile process where I work. Though I find that more testing is better, doing three rounds of testing, usually with about three people in each round, is part of our general policy for building a product. The number of tests is the same whether for a new product or the redesign of an existing product, but the type of users may be different with each. When doing a redesign, it is often easy to find users of different levels that can give valuable input. Expert users are brought into the process very early in focus groups and brainstorm sessions. A low fidelity prototype is presented to these experienced users in order to flesh out more "use cases" or user needs. Expert users are the ones that can tell you what they really loathe, day in and day out and what takes them the most time. These are opportunities for big wins, especially if the product is heavily used internally. These wins save the company money directly and give the employees a higher level of job satisfaction. This is not hands-on testing, yet, however, but merely a way to get back to the drawing board with a better understanding of the product. Product managers may have expertise on a product that the designer probably doesn't without doing loads of discovery.
The first round of usability testing begins after the low fidelity prototype has been embellished to include all the new use cases discovered in the focus groups and brainstorming sessions. I use Flashbuilder to build the prototypes which I label as "medium" fidelity since I see a high fidelity prototype as one that has real code and data behind it. My medium fidelity prototypes are clickable with static data and generally no visual design. Expert users are great for the first round of usability testing, or prototype testing as I tend to call it. Expert users can give you specific usability information but also a higher level of confidence that the IA has been done well and that the interface is actually an improvement over the current product. I tend to do most major changes between and after the first 4 or 5 tests with knowledgeable users. Expert users also tend to get the greatest satisfaction out of helping build a new product. They are really the ones that know the product and including them in the process just builds interdepartmental cohesion. Experts may not expose all your usability problems and heuristic neglect, but they will definitely set you in the right direction.