They are using serverless architecture. Firestore is a "Daas". The latency for loading the page wasn't the real problem, it was the side-effect. The real probl

They are using serverless architecture. Firestore is a "Daas". The latency for loading the page wasn't the real problem, it was the side-effect. The real problem resulted in 40bilion request, they were requesting all the collection of payments in each session. It reaches 16K documents downloaded /session, so they were paying 16K reads documents for session. Firestore gives you 50K reads/day for free, and ask $0.06 for 100K reads. So It was a really bad logical approaching from them. If they did in a best way , like update a document that holds the two accumulators : total money accumulated and total payments created, and read just this document in each session request their bill to show this information for 2M session should be less than $2, because you are reading 1 document for session instead of 16K documents. And pay attention that their system wasn't down, probably the 30s latency was a side-effect of downloading 16K document of each payment information in the client side.