Docs: --lease keeps working (only slower) while a lease's worker is starting #991

What happened

#405 (PR #989) documented that, while a lease's worker is starting, a remote simlock simctl|adb command without --lease fails with "No lease is held by …". The docs do not say what happens with --lease.

With --lease <id>, the CLI skips the lease lookup (src/cli/commands/passthrough.ts:56) and sends the command straight to the lease. The gateway's lease index still holds the lease, so it forwards the command, and the worker parks it until its startup checks finish (docs/internal/flows/startup.md:43). The command runs, only later, by up to about a minute.

What you expected

website/src/content/docs/docs/using-devices/working-with-a-device.mdx (the "Through a gateway" list, near line 89) says that passing --lease <id> avoids the error: the command waits for the worker and then runs.

How to reproduce

Read the page after #989. It mentions only the error case.

Environment

main after b40ff84. Found in the #989 review (non-blocking note). Docs only.

Logs and evidence

Maintainer decision: add the one sentence. Verify the forwarding path against the code before writing it (gateway device.exec through the lease index while the worker is starting).

Written by an agent.