My impression is that it really shines when you need to scale and you can easily spin up and kill large amounts of nodes/compute without any impact to your serv

My impression is that it really shines when you need to scale and you can easily spin up and kill large amounts of nodes/compute without any impact to your service–when you're at the point where you're thinking about the health of the service rather than individual nodes. The whole pets vs. cattle ideal seems to be discussed most in configuration management contexts (because it's widely applicable to most architectures) but it translates into big benefits when operating infrastructure for something like a scaleable SOA. So to give them the benefit of the doubt, they may have made some assumptions about your scale/workload/architecture. Perhaps for them containers occasionally crashing is just a small blip that will be automatically corrected so while they need to be aware of it to monitor for trends they're not generally concerned with them.