They deemed getting to the market fast was more important than hardware costs. If their architecture is good, then they were probably really quick and flexible using Python to glue their architecture together which would have been the development bottle neck to getting the service up and running. All the components of the service can be done by separate teams in whatever language they deem most effective.