Following the oc adm upgrade --to-multi-arch rollout, the cluster's core components were disrupted for an extended period of time, leading to service downtime.
    

RHOCP 4.16.35

Between, roughly, 01:30 UTC and 04:00 UTC, etcd:
- Database back-end size was seen to increase significantly, peaking at close to 7,5 GiB @ 3:45 UTC... from the regular 4.8 GiB in size (before and after the peak)
- Went to multiple iterations of outage and leadership changes.

RAM consumption was reasonably low, averaging 20-30% and only with with peaks ~60%, but CPU showed a linear increase from 200% @ 01:44 UTC to 413% @ 03:56 UTC

- Between 01:30 and 04:00 UTC the kube-apiserver was seen to be recurrently unavailable.

For multiarch to not lead to service disruption.prevent 

The environment is quite large in size, which may have potentially contributed to the sudden resource consumption peak.