How much money would be spent on the scalability testing to see if the service can handle an Apple WWDC without crippling the rest of Twitter?
I think you're greatly underestimating what they did with Summize. It's not rocket science, but there's real (and tried-by-fire) value there for Twitter. Summize didn't get $750k in angel funding to sell to someone in low 5 digits.
We're talking about live search over a medium completely made up of small, frequent text messages. Real-time searching increases Twitter's value as the instantaneous information source it has become. They don't have a lot of strengths to build on, but that's definitely one of the most valuable things Twitter has to offer and Summize augments that.
No brainer: buy Summize for a 2-4 million, add their search tools, get some of their engineers helping in the scalability problems, improve the API based on their feedback, have some of them continue working on their lab prototypes, and keep growing.
What do you think it costs to hire a good engineer? How about a good engineer who is already pretty intimately familiar with your offering's API? What about a collection of good engineers that are already used to working together?