For the past couple of days I ran a test on a tier 3 all-in-one project. The setup of the test projects was exactly the same bit for bit, except for the email settings which in one case were a batch of 10 hotmail accounts and in the other a set of 10 rediff emails.
The hotmail accounts outperformed the rediff ones by about 10 pp (percentage points, so e.g. 50%+10pp = 60%, but 50%+20% = 60%) (i.e. roughly the same amount of submissions, hotmail gave me +10 pp verifieds). The result based on the sample size had a 98% confidence level (i.e. there is a 98% chance the hotmail accounts will perform much better than the rediff ones).
Interested in a batch of 20 yahoo emails for testing purposes now - if you have them, just pm me.