How I built and scaled real‑time chat app from 2 to millions of users — without a rewrite?

The HTML over the wire part stood out to me. Fewer moving pieces usually means fewer weird bugs, especially when you're building and maintaining everything yourself.

One thing I really liked here is that the architecture optimizes for failure scenarios, not just happy-path throughput. The transactional outbox + CDC pattern solves the reliability problem first, and the scalability benefits naturally follow.

Another takeaway is your point about keeping the app tier stateless by pushing fanout downstream. That's a design choice that pays off long before you hit "millions of users" because deployments, scaling, and operations all become simpler.

Also appreciated the discussion around HTML-over-the-wire. Whether someone agrees with that choice or not, it's refreshing to see the trade-offs explained instead of treating any architecture as universally correct.

Great write-up that focuses on engineering decisions rather than just technology choices.