Staging servers have non-deterministic failures

[larsyencken]

Problem

Anecdotally, it seems like there's often a random failure in a staging server pipeline, but if you manually retry it on Buildkite it works.

Expected behaviour

We'd hope for things to work the first time

Gathering more information

• Could we periodically review recent builds on each branch?
• Could we dump all pipelines including failures and retries using the API?
• Are there additional steps that should have automatic retries