fly scaling optimization
Ha! I’d prefer to simply use fly’s built-in autoscaling (if not using experimental features such as multi-process and mounts) once I’ve profiled how many reqs (preferably mimicking a real world pattern) my nodejs app can service before it breaks.
Depending on eng budget and timelines, it’d not hurt to bake in caching / workload isolation / admission control / load shedding / traffic shaping of some sort in the nodejs app if you’re expecting thundering herds. Easier said than done, of course.