We use a nightly build and test environment with TestComplete. The ProjectSuite test run takes 12h and has a huge test result tree. It is pretty complex to compare the actual ProjectSuite test resu...
I think one danger of working with such "diff" reports is that what if the same test is run but with for example 80% of the scriptcode is disabled, in the second run? Then, the numbers on the diff reports don't make any sense!
I think this is only workable if you have a solid not-changing test framework in place, in which you do not (or very very limited) make any changes in the testlogic/scripts itself.