Given that Google does rewrite the code of many companies that they buy, it wouldn't be surprising if they did rewrite the ITA code.
I imagine data persistence and distributed workloads are handled by the C++ portion of ITA code, which means plugging into GFS, BigTable, Chubby and more wouldn't be that difficult. ITA had already done the hard job of working around Common Lisp's missing pieces (a standard on concurrency and systems/network programming), something which would have been a huge distraction to e.g., Google in 2001 (when Ron Garrett attempted to evangelize Lisp).
Lastly, ITA isn't dependent on an acquisition by Google in the same way a start-up with no revenue (being acquired for talent and/or audience) would be. They're being bought partly for talent, but also they're being bought for actual technology and a working, profitable business. That gives them a lot more leverage in setting the terms of the acquisition (has e.g., Postini undergone a full re-write?). These issues is what "technical due diligence" is supposed to work out and resolve: would an integration necessarily mean driving the employees away? If this is so and such integration would have to occur, then the merger may simply not happen.
Even in the case of a pure talent acquisition care has to be taken to avoid an exodus. Money isn't going to the deciding factor here: I'd imagine ITA employees are less bound by golden handcuffs; they're not staring at life changing sums of money and (like most hackers) are more motivated by doing interesting work with smart peers and powerful tools, than by what is effectively a equivalent to a yearly bonus (I am assuming they're likely looking at $100,000-$200,000 in shares vesting over a four year period).