3008.x: setting worker_pools_enabled=False crashes zmq_device via RequestRouter

Description

On 3008.x, setting worker_pools_enabled: False in the master config to fall back to the legacy single-queue routing path (salt.transport.zeromq.RequestServer.zmq_device) causes the RequestServer.zmq_device process to crash in a tight restart loop, rendering the master unable to answer any minion requests.

Root cause

zmq_device unconditionally constructs a salt.master.RequestRouter (salt/transport/zeromq.py:602):

router = salt.master.RequestRouter(
    self.opts, secrets=secrets or getattr(self, "secrets", None)
)

RequestRouter._build_routing_table (salt/master.py:1612) then validates the worker_pools config and raises:

ValueError: Worker pool configuration must have exactly one pool with catchall ('*') in its commands.

because worker_pools_enabled: False causes get_worker_pools_config to return None without leaving a valid worker_pools layout for the router constructor.

The legacy zmq_device path is the ONLY path exercised when pools are disabled; it should not depend on a valid pool configuration.

Reproduction

# /etc/salt/master
worker_pools_enabled: False

Then start the master. Log fills with:

[ERROR   ][10727] An un-handled exception from the multiprocessing process 'RequestServer.zmq_device' was caught:
  File ".../salt/transport/zeromq.py", line 602, in zmq_device
    router = salt.master.RequestRouter(...)
  File ".../salt/master.py", line 1577, in __init__
    self._build_routing_table()
  File ".../salt/master.py", line 1612, in _build_routing_table
    raise ValueError(
ValueError: Worker pool configuration must have exactly one pool with catchall ('*') in its commands.

ProcessManager restarts the crashed zmq_device every 10s indefinitely. salt '*' test.ping times out.

Workaround

Also add an explicit worker_pools catchall alongside worker_pools_enabled: False:

worker_pools_enabled: False
worker_pools:
  default:
    worker_count: 3
    commands: ["*"]

Suggested fix

Either

1. zmq_device skips RequestRouter construction when not self.opts.get("worker_pools_enabled", True), or
2. RequestRouter.__init__ returns a no-op router (or is not constructed) when pools are disabled.

Version: 3008.2+152.g4a10864f6a (CI RPM from PR #69908).

Start in salt/transport/zeromq.py at RequestServer.zmq_device and trace its RequestRouter construction into salt/master.py, especially _build_routing_table. Reproduce with worker_pools_enabled: False, then verify the master stays running and responds to salt '*' test.ping without requiring an explicit worker_pools catchall.