Great post! I’ve been doing some Lambda internals research and can add some detail on the freezing/thawing process: it’s using the Linux kernel’s cgroup freezer subsystem (https://www.kernel.org/doc/Documentation/cgroup-v1/freezer-subsystem.txt).
Also did some looking into (using /proc) when new containers get created for concurrent requests—it’s pretty interesting. Sometimes the container for a concurrent request gets created on the same EC2 host, sometimes another. You can also see the container cycle through different hosts and availability zones when the Lambda EC2 host itself gets retired. Did a talk about it a couple weeks ago, more details/slides here: https://speakerdeck.com/smithclay/faas-measurement-fundamentals