What Happens If TradingView or Your Exchange Goes Down?
Failover Design for Automated Strategies

"Every exchange has some kind of disconnect protection, right?" They do — but you may not qualify. Bybit's dead man's switch (officially, Disconnect Cancel All) is blunt about it in the docs: institutional clients only, and VIP clients cannot access it. If you're a retail trader automating on Bybit, that safety net simply cannot be turned on.

This post is about gaps like that: the moment something upstream breaks, which protections genuinely exist and which only exist in your head. Short version — your open position doesn't vanish because you lost connection, but your stop-loss might effectively not exist at all.

Where your stop-loss lives decides what happens the moment you go dark

There are two ways to place a stop. Both get called a stop-loss, and their fates during an outage are opposites. Get this straight before anything else.

This is why TVSBot places take-profit/stop-loss as real bracket orders directly on the exchange for futures positions, rather than only tracking a level in software — a resting order survives your connection dropping; logic that never reaches the exchange does not.

The exchange's auto-cancel-on-disconnect isn't for everyone

A dead man's switch works like this: you keep pinging an endpoint as a heartbeat, and if it stops for longer than a countdown you set, the exchange automatically cancels your resting orders. Eligibility varies a lot across the four.

Bitget puts its maintenance window in the API docs. The others don't.

Every exchange schedules maintenance. What differs is how precisely the official docs will tell you when. These three say something concrete:

For Binance we found no independent public status or incident page for binance.com, only a GET /sapi/v1/system/status endpoint reporting normal or maintenance. Which makes Bitget the only one that tells you, while you're still writing the code, not to schedule anything important on Tuesday, Wednesday or Thursday afternoon.

What TradingView's side actually admits to

TradingView runs a public status page at status.tradingview.com, with a dedicated "Alerts" component. As of this check it showed operational, and the only alert-related incident logged recently was "Technical alerts partly unavailable" (Jul 1, 10:14 – Jul 3, 18:23 GMT, marked resolved). Outages happen and are tracked publicly — but TradingView has never published whether alerts that should have fired during one get resent afterwards.

Worth being precise here, because a well-documented webhook resend mechanism often gets assumed to cover this. It only applies when your own receiving server errors: on a 5xx (with 504 explicitly excepted in TradingView's own wording) it resends after 5 seconds, up to 3 times; on a 504, a 4xx, or no response within 3 seconds there is no retry — it's treated as delivered and dropped. TradingView's own service degrading is a different scenario. For the receiving-end breakdown see our webhook delay deep-dive; if alerts aren't firing at all, this troubleshooting checklist covers the more common causes first.

The sentence we can't quote verbatim

No exchange states in plain language that "your position will not be closed if your connection drops." That conclusion is our own synthesis: every dead man's switch document we found scopes itself to orders, never positions, and liquidation risk is managed by a separate margin engine that doesn't reference your connection status. It's a reasonable inference backed by several independent official sources using the same careful wording — but it is an inference, not a direct quote, and we want to be upfront about the difference.

Questions to ask yourself before you go live

The ones you can't answer are your gaps:

• Is my stop a real order resting at the exchange, or does it only exist inside strategy logic?
• Does my account tier actually get this exchange's dead man's switch? — if you're not sure, assume no
• For the short disconnects exchanges say won't be announced (Bybit: under 10 seconds), does my bot reconnect and retry by itself?
• After an incident, can I tell whether the alert never arrived or arrived and failed downstream? — only a signal log kept in advance answers that
• If the thing that goes down is my dashboard, do I have a second channel to find out, and a way to close everything manually?

Being honest: no design guarantees zero missed signals

No automated trading setup can promise 100% delivery — ours included. TradingView has never published an SLA for an alert missed during its own outages; we don't publish an uptime guarantee for TVSBot either, because we don't have a public SLA to point to. Design reduces the odds and the blast radius of a missed signal; it doesn't eliminate the possibility. Anyone claiming otherwise about their own platform is telling you something they can't back up.

FAQ