Send order when A/B testing
As far as I can see current behaviour when processing an A/B test mailing is that CiviMail sends out to all A recipients then all B recipients.
If you have rate limiting on outbound email that is significant relative to your list size, this can result in a significant difference in (average) send time between the A group and the B group, which may then be confounding your A/B test results.
I wonder how hard it would be to implement some kind of interpolation (?) for the sends for an A/B mailing -- something like firing off equal numbers of As and Bs in each mailing batch as they are processed?