TL;DR: Please implement a basic horizontal autoscaler built into railway that monitors at least CPU usage and increases or decreases the replica count based on the aggregated CPU% of all

I've seen this topic proposed in the past but for some reason it never gets any attention and I haven't found any reason as to why its not implemented. Other PaaS such as Render and Heroku already handle elastic loads by monitoring the aggregated CPU usage for a given service and scaling up or down the replica count.

The lack of any container autoscaling system is preventing us from fully migrating off GCP for some services that handle very variable load and must be able to scale quickly without manual intervention (I've seen Railway state that you can always update the replica count in the dashboard which is... less than ideal in a production environment)

Of course, CPU% scaling has its problems and other methods such as queue time but that would require manual instrumentation using a custom library, although there are paid services like JudoScale that do this (for a hefty price, of course)

For now, I've developed a simple service that is deployed alongside the container I want to scale and it monitors its CPU usage to determine when to use the Railway API to increase or decrease the replica count.