on top of #120, it would be useful if the health check also performed some smoke tests like

• has there been a successful delete in the last minute? → is there at least one account with last_delete sometime within the last minute?
• same with fetch, refresh
• is there an abnormal amount of accounts with high backoff levels? this one would need additional monitoring to measure what is "normal" as mastodon instance churn likely means it keeps going up. I'm not sure it's worth it
• others?

probably would be wise to only trip these tests if there are more than a few accounts, so that a new install doesn't keep raising flags just because there aren't enough accounts to keep it busy