Noisy neighbor starves pod

Your multi-tenant SaaS runs many customers' workloads on a shared Kubernetes cluster. This afternoon several unrelated tenants simultaneously report slow API responses and a few 503s. Your own service's request rate is normal, CPU on its pods is fine, but node-level dashboards show one node's CPU and disk I/O pegged, and the affected pods all happen to be scheduled on or near that node. One tenant just kicked off a heavy export/backfill. How do you triage and mitigate, and what's the durable fix?