Hi Alistair (et all),
while it's true that still there are a number of issues which code has still not landed, this is something we were discussing/evaluating last week before deciding to proceed with the 3.1 QA round.
Basically what we have done is:
1) Whenever the issues are global (not about a given plugin) (CBE issues are the main ones of this type), we have preventively failed related existing QA tests, so nobody can try them. Once the feature is complete, they will be moved back to available.
2) When the issues are confined within the limits of a plugin (say mod_feedback, or mod_lti as 2 examples of plugins having pending issues), we have allowed related QA tests to be tried. The idea is that old described behaviors should continue working in a BC way. Also, of course, if we have time... the idea is to reset them once we are feature complete, so they are re-run with the final code.
3) Apart from the two points above, that are about existing QA tests... there is a number of new tests that are being created (right now) and have not been still added to the 3.1 QA round. They describe new behaviors/features and only will be tested when we are feature complete.
Basically we were discussing if was better to delay QA start or proceed with it ASAP, knowing that some test will need to be rerun (2) and will be added later (3). And looking forward to get possible problems detected earlier (and have time to fix them), we decided that starting ASAP was the best choice, with the exception of CBE (really transversal) being preventively failed to avoid problems till completed (1).
Let's see how it goes, sure we learn from it, thanks for your collaboration!
PS: Note I maybe slightly wrong/inaccurate with some of the points above, although I think that was, overall, the spirit of the decision. Sure that if any detail is wrong, somebody from HQ will clarify it soon.