Good call — that's exactly right, and worth calling out. I didn't want people to read "split by domain" and think none of this helps until they do a full rearchitecture. Pooling, a read replica, and caching are honestly the easiest wins precisely because they bolt onto a monolith with zero architecture changes — a couple days of work that can save you on launch day, way before you're ready for the bigger split.
Ale
I'm a software engineer, founder and maker. I like to work with cutting edge technologies and build new products.
The point about workload contention hitting right during launch day is painfully accurate. Running heavy analytics queries or background jobs on the same cluster serving live user logins is a silent killer. Even within a monolith, introducing basic connection pool discipline, read replicas, and a caching layer before your first traffic spike is the difference between surviving a launch and taking down your entire app.