> If Heroku wasn't providing enough performance for their money (regardless of technical cause), then why did they stick around so long?

We didn't know how bad the performance was because Heroku's tools (logs and New Relic) reported incorrect performance data (i.e., they said requests weren't queuing when they were actually spending a ton of time queuing)

EDIT:

> Would Rap Genius still have sued if the documentation had been 100% correct and instead the problem was just plain old slow I/O on Heroku's side?

To be clear, we have no official association with http://herokuclassaction.com/ or the lawyer behind it – I actually found out about the site from the article's author when he interviewed me for the story.

(But I do think Heroku owes its customers a refund)