Sequential scan on large tenant

A B2B SaaS app was fine for a year, then this week, after onboarding a large enterprise customer (tenant_id 4471 with 20M rows vs the typical 50k), the dashboard-load endpoint times out, but only for that tenant. Other tenants are snappy. DB CPU and I/O climb noticeably whenever 4471 loads a dashboard. EXPLAIN ANALYZE of the dashboard query shows a sequential scan filtering WHERE tenant_id = 4471 AND created_at > now() - interval '30 days' over a 60M-row events table; there's an index on tenant_id alone but none on (tenant_id, created_at). Triage and fix.