I agree with your points, especially on the topic of random procs and so forth. A true DPS test in-game will require a very large sample of data to be accurate (especially when you want it VERY accurate) and can require you to find 5 guys to buff you all the time. Often, in-game DPS tests are done poorly, resulting in invalid data.

The key reason behind my statement that in-game testing is more important than TC'ing is that I don't honestly believe there is a 100% accurate spreadsheet out there, as Blizzard never reveals every aspect of gameplay and mechanics to us. Testing damage on a live realm (with the proper testing methods, sample size, etc) will ensure that any small change that Blizzard has made will not go unnoticed. The sheets are very very accurate, but running these numbers through Blizzard's formulae is the truest test (if done properly).

Thank you for your comment, and reminding me that it's important to know that math doesn't lie, but poor testing can. :)

Your point about not having a spreadsheet that is 100% accurate to Blizzard's own calculations is a good one. For all that has been done, calculated, theorized and even occasionally blue-confirmed, there are still some areas where we're still working in a black box.

I would only add that the mechanical work done on the EJ rogue spreadsheets is largely the result of exactly the kind of extensive in-game testing that you describe.