Standing up a side service is a five minute job instead of a Friday. We push the repo, it picks up the Dockerfile, and private networking between services works without me hand writing a VPC. For a two person team that's the whole pitch.
What needs improvement
The usage based billing is the part that bites. A runaway container is something you find out about on the invoice rather than something that gets refused at deploy time, and I'd trade a lot of flexibility for a hard spend cap that just stops. Logs also get thin fast when you actually need to go back a few days.
