[Bug] Server restarts in incorrect state when using file-backed persistence

[dandavison]

Describe the bug

When using file-backed persistence, restarting the server after a ctrl-c shutdown brings it back up in an incorrect state.

Minimal Reproduction

1. temporal server start-dev --db-filename=/tmp/db
2. ctrl-c once it's started
3. temporal server start-dev --db-filename=/tmp/db
4. You'll now get a stream of these errors (or if not, try ctrl-c again and start server again)

time=2024-08-09T11:56:02.240 level=ERROR msg="service failures" operation=PollWorkflowTaskQueue wf-namespace=temporal-system error="Not enough hosts to serve the request"
time=2024-08-09T11:56:02.243 level=ERROR msg="Unable to call matching.PollActivityTaskQueue." service=frontend wf-task-queue-name=/_sys/temporal-sys-processor-parent-close-policy/2 timeout=1m8.90395s error="Not enough hosts to serve the request"

5. The main UI root page is served correctly: http://localhost:8233/namespaces/default/workflows
6. If you try to start a workflow that works, but then navigating to the workflow page is 503 Not enough hosts to serve the request

Environment/Versions

1.0 and when built against recent server.