It's easy from an armchair to say things like they should just get more capacity, or they should segregate free repos onto other servers. But maybe they are doing what you're thinking, and with that churn they get mistakes that cause outages, esp dealing with legacy systems. I've been there (not at GitHub or MSFT).