in all articles about railgun (include this) I can not find anything about how to do a horizontal scale.
I worry about those 2 points:
what happens if the railgun machine goes down. Will all the users get a 5xx errors?
how many QPS can one railgun node hold, I know it depends on the machine’s hardware and the content of the response, but I want to know if the user requests count raise up suddenly, will railgun slow down the request in turn?