Describe the bug
Running authentik 2025.8.0 on Kubernetes with IPv6 disabled causes the worker’s metrics and healthcheck processes to crash at startup. The processes try to bind a dual-stack IPv6 socket via socket.create_server(..., dualstack_ipv6=True), and Python raises:

ValueError: dualstack_ipv6 not supported on this platform

This error appears in both MetricsMiddleware.run(addr, port) and WorkerHealthcheckMiddleware.run(host, port) logs. This prevents worker pods from starting cleanly.

To Reproduce

1. Deploy authentik 2025.8.0 in a Kubernetes cluster where IPv6 is disabled at the node/pod level.
2. Start the worker pods.
3. Observe worker logs; the worker metrics or healthcheck subprocess crashes with a traceback ending in ValueError: dualstack_ipv6 not supported on this platform.

Expected behavior
The worker metrics and healthcheck HTTP servers should fallback to IPv4-only binds on IPv4-only systems, or respect the configured metrics listener (e.g. AUTHENTIK_LISTEN__METRICS=0.0.0.0:9300) without attempting a dual-stack IPv6 bind. The worker pods should start cleanly without crashing.

Logs
Example traceback from worker pod:

  File "/authentik/tasks/forks.py", line 43, in worker_metrics
    MetricsMiddleware.run(addr, port)
  File "/ak-root/packages/django-dramatiq-postgres/django_dramatiq_postgres/middleware.py", line 282, in run
    server = HTTPServer((addr, port), _MetricsHandler)
  File "/usr/local/lib/python3.13/socketserver.py", line 457, in __init__
    self.server_bind()
  File "/ak-root/packages/django-dramatiq-postgres/django_dramatiq_postgres/middleware.py", line 42, in server_bind
    self.socket = socket.create_server(self.server_address, family=self.address_family, dualstack_ipv6=self.address_family == socket.AF_INET6)
  File "/usr/local/lib/python3.13/socket.py", line 912, in create_server
    raise ValueError("dualstack_ipv6 not supported on this platform")
ValueError: dualstack_ipv6 not supported on this platform

Similar errors appear for worker_healthcheck.

Version and Deployment

• authentik version: 2025.8.0
• Deployment method: Kubernetes (Helm)
• Kubernetes cluster: IPv6 disabled on nodes and pods

Additional context
We attempted several workarounds without success:

• Set AUTHENTIK_LISTEN__METRICS=0.0.0.0:9300 on both server and worker.
• Set dramatiq_prom_host=0.0.0.0 and dramatiq_prom_port=9191.
• Set PROMETHEUS_MULTIPROC_DIR=/dev/shm/authentik_prometheus_tmp.
• Enabled IPv6 sysctls inside the worker pod.
  None of these prevented the crash.

There is an existing issue (#16292) about an empty metrics response, but our case involves the socket bind. A temporary workaround is to enable IPv6 on the host, but a fix to gracefully handle IPv4-only environments would be appreciated.