[aenis]

Sounds great (nearly too-good-to-be-true great). Wonder how the SLA will look like. I have been using gcs, s3 and firestore - and their actual reliability varies significantly, while advertised slas are similar. For instance, with firestore one has to implement a pretty lenient expotential backoff in case of a timeout, and if the backoff results in the object being retrieved in, say, 2 minutes -- thats still ok as per gcs sla. It obviously makes it hard to use it for user-facing stuff, such as chatbots, where you can't afford to wait that long. In my anecdotal experience of using firestore for about 10 million operations per day, we will usually have a problem like that every few days, and that means user-noticeable failure. It would be great to read more on cloudflare's approach to reliability defined as "99%" percentile max latencuy. Can't wait to give it a try with our workloads.