Make storage administration resilient to a UI Pod loss #5694

🤖 Generated by the Agentic Engineer

Outcome and evidence

Storage administration should remain available when one UI Pod is replaced. A current read-only check agrees with the declared single-replica setting and Coroot's availability warning. The UI uses temporary volumes; the retained storage data plane is a separate responsibility. This is an availability risk, not evidence of lost volumes or a current storage outage.

Acceptance

• Declare at least two production UI replicas through the existing configuration owner, preserving the chart/image, authentication, storage and network boundaries.
• Validate the actual rendered chart and effective Pod labels, placement and disruption behavior. Prove the replicas occupy different eligible nodes; a replica count alone is insufficient.
• Use the normal protected rollout and verify both current Pod identities, readiness, endpoint membership and the authenticated administration path afterward.
• Observe the availability warning retire naturally. Do not dismiss it, change its threshold or induce a destructive node/storage trial for evidence.

Planning

Bug; Medium priority, Low effort, Standard service class. Capture in Backlog while existing delivery work and capacity gates remain in effect. No implementation or additional draft allowance is implied.

Part of #4981.