Completa il hardening autoscaler (Fase 4): budget, kill-switch e runbook.

Drain sicuro in scale-in, alert Ops su overflow e controlli admin senza abilitare il deploy in produzione.

Co-authored-by: Cursor <cursoragent@cursor.com>

backend/app/services/ops/health_checks.rb

@@ -44,7 +44,8 @@ module Ops
         check_sidekiq_heartbeat,
         check_sidekiq_dead,
         check_http_rails,
-        check_rails_latency
+        check_rails_latency,
+        check_stream_overflow
       ]
       findings << check_http_public if public_check_due?
       findings
@@ -161,6 +162,58 @@ module Ops
       fail_finding("garage_storage", "warning", "garage_storage:head", "Garage storage non raggiungibile", e.message)
     end
 
+    def check_stream_overflow
+      return ok_finding("stream_overflow:skip", "Nodi stream non migrati") unless ActiveRecord::Base.connection.data_source_exists?("stream_nodes")
+
+      orphan_hours = ENV.fetch("STREAM_OVERFLOW_ORPHAN_HOURS", "3").to_i
+      orphans = StreamNode.where.not(slug: "home").where(status: %w[ready draining]).select do |n|
+        n.active_publishers.zero? && n.created_at < orphan_hours.hours.ago
+      end
+      metrics = Streams::Autoscaler.metrics
+      over_budget = !metrics[:within_budget]
+      at_max = metrics[:overflow_nodes] >= metrics[:max_overflow_nodes] && metrics[:free_slots] <= metrics[:soft_free_slots]
+
+      if orphans.any?
+        return Finding.new(
+          kind: "stream_overflow",
+          severity: "warning",
+          healthy: false,
+          title: "Nodi stream overflow idle",
+          message: "#{orphans.size} nodo/i idle da >#{orphan_hours}h: #{orphans.map(&:slug).join(', ')}",
+          metadata: { "slugs" => orphans.map(&:slug) },
+          fingerprint: "stream_overflow:orphan_idle"
+        )
+      end
+
+      if over_budget
+        return Finding.new(
+          kind: "stream_overflow",
+          severity: "warning",
+          healthy: false,
+          title: "Budget overflow streaming",
+          message: "Stima €#{metrics[:estimated_monthly_eur]}/mese > budget €#{metrics[:monthly_budget_eur]}",
+          metadata: metrics.transform_keys(&:to_s),
+          fingerprint: "stream_overflow:budget"
+        )
+      end
+
+      if at_max
+        return Finding.new(
+          kind: "stream_overflow",
+          severity: "warning",
+          healthy: false,
+          title: "Capacità stream al massimo",
+          message: "overflow=#{metrics[:overflow_nodes]}/#{metrics[:max_overflow_nodes]} free_slots=#{metrics[:free_slots]}",
+          metadata: metrics.transform_keys(&:to_s),
+          fingerprint: "stream_overflow:at_max"
+        )
+      end
+
+      ok_finding("stream_overflow:ok", "Overflow streaming OK")
+    rescue StandardError => e
+      fail_finding("stream_overflow", "warning", "stream_overflow:error", "Check overflow fallito", e.message)
+    end
+
     def check_sidekiq_heartbeat
       redis = Redis.new(url: ENV.fetch("REDIS_URL", "redis://localhost:6379/0"))
       last = redis.get(Ops::HealthMonitorJob::HEARTBEAT_KEY).to_i