Inhibit executing a query asap for 5 minutes if the same query was rejected due to busting limits

[krajorama]

Is your feature request related to a problem? Please describe.

If a query loads too much data (chunks/samples/series) and we rejected, nothing stops the user from issuing the same query again right away. This is not a good strategy as we can be reasonably sure that we cannot execute the query a second time around and generate useless load.

Describe the solution you'd like

Reject repeated queries as soon as possible for 5 minutes if the query failed due to limits and has the same query and same/similar time range.

Describe alternatives you've considered

N/A

Additional context

N/A

[56quarters]

Sounds like negative caching for the results cache in the query-frontend.

[pracucci]

Looks a duplicate of https://github.com/grafana/mimir/issues/2676.