What is the problem you are trying to solve?

We currently cache non-transient errors per #9028. However this is only applied to range queries. This becomes an issue with badly written recording rules (e.g. rules that have huge range selectors): they'll be run every minute and hit limit errors every time, creating a huge amount of load for no reason.

Which solution do you envision (roughly)?

We should extend the error caching middleware to support instant queries. This would prevent repeated evaluations from creating unnecessary work to some extent (we cache errors for 5m by default).

Have you considered any alternatives?

No response

Any additional context to share?

No response

How long do you think this would take to be developed?

Small (<= 1 month dev)

What are the documentation dependencies?

No response

Proposer?

No response