Free egress is not free serving. R2 still bills every read, S3 does not bill origin transfer to CloudFront, and the variable that decides both bills is cache hit ratio. An evenhanded comparison of Cloudflare R2 vs Amazon S3 for media offloading, with the four levers that move the number and a decision procedure you can run against your own traffic.
Third-party API spend is the one production signal with no error rate attached to it, which is why it creeps up quietly. This is a working engineer's guide to reducing SaaS API costs by changing the shape of your calls: reading the billing unit before you optimise anything, killing pointless polling with conditional requests and webhooks, collapsing N+1 patterns, caching with stampede protection and per-tenant keys, stopping your own retry amplification, and attributing spend so you can prove the work paid off.
Put a CDN in front of a dynamic site carelessly and it will cache one user's personalised page and serve it to everyone — invisibly, because your logged-in browser looks fine. Here is how to classify routes, set cache headers that hold, and run the two curl commands that catch the leak.