Fix instability of CI tests

This is mostly affecting:

• make tests-refresh and
• make tests-refresh-resilience

For now, we're setting these CI jobs to be manual (in TBD), since they pass reliably locally.

Possible causes/solutions:

• In some tests we run rugged initialize --local. We suspect that it is taking some time for the repo metadata and keys to sync to the workers (volume mounts, NFS, etc.) We might want to consider running without the --local option.
• Alternatively, we might want to update the initialize command, to have it wait for NFS to sync, or something.
• Waiting for a specific amount of time does not seem to be reliable. Maybe we should add a step definition that'd look like: And I wait for "STRING" in "FILE" (with a timeout, etc.)