Some shards have moderate storage space pressure from test instances where users, e.g., imported a large repository. Although we do not backup instances that have been out of service for several days, these instances still consume storage space on primary data partitions.

We should have a proper destruction process for instances, and automatically destroy test instances some time after they are suspended. We need to:

• Destroy the repositories.
• Destroy backups (although probably fine to leave these for now, they'll cycle out over time and we'd never normally destroy an instance that hadn't been suspended for long enough for the backups to cycle).
• Destroy the S3 bucket (maybe just deal with this separately for now, since other administrative actions do not have direct S3 access).
• Destroy the databases with bin/storage destroy.