I don't think that's a fair characterization of what happened. First, they scaled a simple solution for a long time and migrated away from it before it exploded spectacularly. I don't recall them explicitly saying how much time they spent on it, so how is it possibly to say whether it was worth the time or not? I also don't remember if it was here or on Twitter, but one of the engineers listed the many features they built and shipped _instead_ messing with their database. I'd say that was a good business decision - they got more features to their customers sooner and dealt with a scaling issue before it impacted customers.