They talk about remediating issues with email alerts. That is exactly the kind of approach you can take when you are nowhere near Google's scale, instead of building a complex system to ensure all of these remediations can happen automatically without any human intervention. Sounds like they took exactly the right approach of not over-engineering a solution prematurely.
Ironically we kind of relied on e-mail alerts for too long and would kinda sorta break Gmail. While it was probably just the web/IMAP interface or browser/client itself struggling, trying to bulk delete tens of thousands of emails while hundreds more per second were coming in often had incoherent results.
I think I was mainly writing in agreement with the article’s closing:
> What we’re all imploring you to do is to think! And to actually understand the problem you are trying to solve.
Clearly, having HA and reliable transport is reasonable for your use-case but that isn't the same as microservices, hadoop or Kafka.