Over the past few months, we've been using PNUnit for three types of testssmoke, load, and merging tests. Smoke tests are run every time a developer finishes a task, together with regular NUnit test suites. Smoke tests basically cover core product functionality, and have been extended to include more and more test cases.
Each time we create a new release (typically once a week), the "integrator" (one of our team members) merges all the tasks to create a new baseline, then executes all the smoke tests. However, he does so by launching them on different platform combinations, covering all our supported platforms. Merge tests (a large test suite covering lots of branch-merging scenarios) are also executed during integration on different platforms. Finally, PNUnit tests are specifically designed to be scalable. Each test implements simple operations so they can be combined on the testconf script, and easily define load scenarios through the available test machines.
We've also started to use PNUnit on a cluster (about 50 Xeon machines), letting us test software under heavy load conditions. The framework has proved to be very useful on our project, having detected many bugs over the months.
There are still several aspects of the framework that need improvement. For instance, tests are currently only run from the command line, and results are gathered this way. A huge step forward would be to include GUI testing support in the framework. Also, it would be useful to be able to record user events and play them back as with other tools, but with the added power of multiplatform support.