Description of problem:

probe_success that is used by the metric rhods_aggregate_availability is failing. Depending on the actual time you execute the query (threshold), the measure is detected or not, giving a glitch, the measure appeared and disappear when refreshing. This problem is affecting previous release of rhods and is important for many measures, including the SLA. This issue is related to RHODS-4229

Prerequisites (if any, like setup, operators/versions):

RHODS

Steps to Reproduce

1. Go to Observe > Metrics
2. Write the query (min(min_over_time(probe_success[10s])) by (instance) or label_replace(min(min_over_time(probe_success[10s])), "instance", "combined", "instance", ".*")) or rhods_aggregate_availability or min(up{job="Traefik Proxy Metrics"}) 
3. Run the query several times until see the glitch

Actual results:

The Traefik Proxy measure appears and disappear or don't appear at all when refreshing, the rest of the measures behave as expected.

Expected results:

The measure don't disappear and is accurate

Reproducibility (Always/Intermittent/Only Once):

Always

Build Details:

quay.io/repository/modh/rhods-operator-live-catalog:1.13.0-rhods-4229

Workaround:

Additional info:

This misbehavior of the traefic proxy measure through rhods_aggregate_availability metric probably owe to the probe_success metric min(up{job="Traefik Proxy Metrics"})  Depending on the actual time you execute the query (threshold), the downtime is detected or not, giving the glitch that we saw.