What is the bug?

When a ruler fleet is undergoing natural churn, e.g. normal pod evictions, any querying of the ruler for rules at that time leads to returning a 500.

For example, for the following request, GET /prometheus/api/v1/rules?type=alert, the server responds with the following:

{
    "status": "error",
    "data": null,
    "errorType": "server_error",
    "error": "too many unhealthy instances in the ring"
}

How to reproduce it?

Start a normal ruler fleet and terminate one pod gracefully while issuing a command to retrieve rule details.

What did you think would happen?

No 500s

What was your environment?

Infrastructure - Kubernetes v1.30.12
deployment tool - custom
version - 2.16.1

Any additional context to share?

No response