> ## Documentation Index > Fetch the complete documentation index at: https://docs.reactor.inc/llms.txt > Use this file to discover all available pages before exploring further. # FAQ > Frequently asked questions about Reactor ### Why is my session stuck in `waiting`? The `waiting` state means Reactor has accepted your request and is assigning a GPU. This usually takes a few seconds. Once the status reaches `ready`, the WebRTC connection is established and media starts flowing. For more information, see [Connection lifecycle](/concepts/sessions#connection-lifecycle). Longer waits typically mean capacity for the model is briefly saturated. Your session moves to `ready` as soon as a GPU frees up, and you don't pay anything while you wait. If a session stays in `waiting` for an unusually long time, share your [session ID](/concepts/sessions#session-id) in [#help on Discord](https://discord.com/channels/1425218224823337094/1447706712898404645) or email [support@reactor.inc](mailto:support@reactor.inc). ### How do I switch models without re-authenticating? You don't need to. Tokens aren't scoped to a specific model. The `POST /tokens` endpoint takes no model parameter. The model is chosen at connect time via `modelName` on the SDK. The same JWT works across every model your API key has access to for the token's full lifetime (up to 6 hours). For more information, see [Authentication](/authentication). ### What happens if I lose network mid-session? The session keeps running on the GPU and all model state is preserved. You have **30 seconds** to reconnect before the GPU session is automatically terminated. When your network comes back within that window, call `reconnect()` to resume where you left off. For more information, see [Sessions › Disconnecting](/concepts/sessions#disconnecting). The SDK reports unexpected drops through the error handler. If `error.recoverable` is true, wait `error.retry_after` seconds (default 3) and call `reconnect()`. See [Using the SDK › Reconnection](/sdk-reference/using-the-sdk#reconnection) for the exact pattern in JavaScript, React, and Python. The GPU is held for you during the reconnection window, so billing continues until you reconnect or the 30-second timeout expires. For more information, see [Recoverable disconnects](/resources/billing#recoverable-disconnects) and [Rate Limits](/resources/rate-limits). ### Can multiple clients connect to the same session? Yes. A session can serve multiple WebRTC connections at once. Create a session on one client or your backend, then have other clients adopt it by passing its `sessionId` to `connect()`. A client that adopted a session (rather than creating it) tears down only its own connection when it disconnects; the session keeps running for its owner. This is a JavaScript SDK feature (2.12.0+). See [Sessions › Multiple connections per session](/concepts/sessions#multiple-connections-per-session). ### Does Reactor work in mobile browsers? Yes. Reactor works in modern mobile browsers on iOS and Android. ### Where do I find my session ID? The SDK exposes it on the session object as soon as the session is created. For the exact field on each SDK, see [Session ID](/concepts/sessions#session-id). Include it any time you contact support or report a bug. It lets us trace your session through our logs. ### How do I check Reactor's status? Visit [status.reactor.inc](https://status.reactor.inc/) for live uptime and incident history. ### How do I report a bug? Email [support@reactor.inc](mailto:support@reactor.inc). Include your SDK version, [session ID](/concepts/sessions#session-id), reproduction steps, and browser/OS. ### Where can I read more about Reactor's policies? Please view our [Terms of Service](https://www.reactor.inc/legal/terms) and [Privacy Policy](https://www.reactor.inc/legal/privacy) for details. *** ## Next steps Contact the team and get help. Understand how sessions are billed and how to minimize cost. Session quotas, token TTL, and what happens when limits are hit. Connection lifecycle, reconnection, and session IDs.