v0.53.10

Fixed

• core/components/grafana-stack/mimir-values.yaml: promote cortex prd ingester override into upstream defaults. The chart's stock 512Mi memory limit was insufficient for the WAL replay on restart — with hundreds of segments accumulated, the ingester OOMKills mid-replay before /ready returns 200 and the Mimir Application stays at Synced/Progressing indefinitely. cortex prd documented this on 2026-05-01 (corrupted WAL → split-brain → forced node drain); cortex hml reproduced the same failure 2026-05-17. Promoted defaults: replicas: 2, requests.memory: 384Mi, limits.cpu: 500m, limits.memory: 2Gi, topologySpreadConstraints with whenUnsatisfiable: DoNotSchedule (prevents bin-pack of both replicas onto the same node, preserving HA). The grafana namespace ResourceQuota in estabilis-platform-gitops already accommodates this (limits.memory: 32Gi, 18.9Gi free on a typical cluster). Total ingester budget rises from 512Mi to 4Gi (2 replicas × 2Gi).