When your exchange is down and a call goes out

The message arrives at 03:12. Long, entry given, stop given, two targets. You open the app and it shows a spinner where the order ticket should be. Ninety seconds later the spinner is still there, the price on your other screen has moved through the entry, and you still do not know whether the exchange is down, your connection is, or the order you tapped went through twice.

Three different things can produce that screen, and they fail in different ways. Separating them is most of the work, because the response to each is different and only one of them can be prepared for in advance.

Which of the three is actually happening?

Scheduled maintenance is announced ahead of time and is usually per-service rather than all-or-nothing. The venue names a window, names what stops, and often names what keeps running.

An unplanned outage arrives with no notice and no scope. Nobody tells you which parts are affected because at the moment it starts, the venue does not know either.

A degraded venue is the worst of the three, because orders sometimes go through. There is no clear signal to stop trying, so people retry, and retrying is how one intended position becomes two.

The three look identical on a phone at 03:12. They are distinguishable in about thirty seconds if you know where the venue publishes its status, which is the first thing worth setting up while nothing is broken.

What does a scheduled window actually switch off?

Not everything, and not the same things twice. Scope is decided per event, by the venue, and published in that event's notice.

A Binance maintenance notice from March 2020 lists the suspension as covering "deposits, withdrawals, spot trading, margin trading, P2P trading, lending, redemption, as well as asset transfers from sub-accounts, margin accounts, futures accounts, and fiat wallets" — and then states that "trading on Binance Futures will continue unaffected during the maintenance period" (Binance, Notice of Temporary System Maintenance). Spot down, derivatives up.

An OKX notice for April 2021 goes the other way. It describes a ten-minute window in which trading across spot, margin, futures, perpetual and options in the unified account "may be temporarily suspended", while reassuring holders that "This maintenance will not affect your account balance and the use of other functions of the website. Your account security is still guaranteed." (OKX, Perpetual Swap Trading And Unified Account Suspended Due to System Maintenance).

Two dated notices, two venues, two different shapes of outage. Neither is a rule about crypto exchanges in general, and neither predicts what the next window at the same venue will cover. The only reliable statement is the one nobody likes: read the specific notice for the specific window, because scope is a per-event fact that changes.

Why does the venue tell you to act before the window, not during it?

Because during the window there is nothing you can do, and the market does not agree to pause with you.

An earlier OKX notice, for a one-hour perpetual swap suspension in June 2020, is unusually explicit about which functions go dark: "market quote, order placement, order cancellations, order amendments, margin addition and transfer". Then it says what to do about it:

"Please take necessary measures before the downtime, such as adding margin, lowering the leverage, or even closing the positions to avoid potential risks like forced-liquidations caused by market volatility." — OKX, Perpetual Swap Suspension For Temporary System Maintenance

That sentence is the whole article in one line. The same notice promises that "Your account balances will not be impacted and your assets security will not be compromised with the system maintenance" — your coins are safe, and that is a genuinely different question from whether your position survives.

Here is the asymmetry. Your ability to act stops. Nothing else does. An open perpetual position is still marked against an index that keeps updating, still charged or paid funding on the venue's schedule, and still liquidatable by a risk engine that has no interest in whether you can reach the app. Adding margin was possible an hour ago and is not possible now, which is precisely why the notice says to do it beforehand. If the distance between your entry and the point where the engine takes over is not obvious to you, that arithmetic is worked through in Liquidation explained — and the distance shrinks while you hold, whether or not you are watching.

What does an unplanned outage look like from the inside?

There is a public example with a timeline attached. OKX published an incident report for 12 January 2021 describing two separate suspensions of perpetual swap trading within a few hours, the second following a system upgrade after which, in the report's words, "an abnormality occurred in the common components between the push system and the perpetual trading system" (OKX, Jan 12, 2021 Perpetual swap system failure report). Trading was suspended and restored twice, each interruption lasting on the order of twenty to thirty-something minutes.

Two things in that report matter for a follower. First, the durations are short — nobody's account was frozen for a week — and short is still long enough to contain an entry, a stop and a target. Second, the explanation arrived afterwards. During the event, the only available information was that orders were not working.

That is the practical difference between the announced case and the unannounced one. In the announced case you get to make decisions in advance. In the unannounced one, the decisions that protect you are the ones you already made.

Why is a half-working venue worse than a dead one?

Because a dead venue tells you to stop, and a half-working one does not.

When the ticket loads slowly and confirmations do not come back, the natural response is to tap again. If the first request was queued rather than dropped, you now have two orders for a position you sized once. When it clears, you are in at double size on a call whose entry has already moved. That is not a signal problem, and no channel's record will ever show it.

The other failure of a degraded venue is a stale screen. A position panel that will not refresh is not evidence that nothing changed; it is evidence that you cannot see what changed. Acting on numbers you cannot confirm is how a hedge gets placed against a position that has already closed.

If the venue publishes a status page, that is the screen to trust over the app. Kraken runs one that separates announced maintenance from a log of incidents already resolved (Kraken system status). Whether your own venue runs one, and at which address, is the thing to find out while nothing is wrong. Knowing your venue's URL before you need it costs nothing and is the difference between thirty seconds of diagnosis and twenty minutes of guessing.

What does downtime do to the call you were following?

Four things, and each one detaches your account from the channel's record a little further.

The entry you miss. The call resolves without you. If it wins, the channel counts a win, and it is not lying — the price did what the post said it would. Your account simply was not there. A missed entry is also not a reason to take the trade twenty minutes later at a worse price, which is its own way to lose money and is covered in when to skip a signal that has already moved.

The stop you meant to move. If your risk plan involved moving a stop to breakeven when the first target hit, and the first target hit during the outage, the stop stayed where it was. Plans that require you to be reachable are only as good as your ability to be reached.

The instruction you cannot obey. "Close now" is not an action available to you during a suspension. The channel will post the exit, that exit will go into its record at the posted price, and your position will still be open at whatever price it reaches when access returns.

The disagreement afterwards. Your result and the channel's result differ. This is normal and boring, and it is worth saying plainly: an outage hour is an ordinary reason for that gap. It is not evidence that a channel is dishonest, and it is not something a channel can be blamed for. Keeping those two categories apart is what makes your own records worth anything later.

What is actually in your control here?

Five things, all of them decided before the screen freezes.

Know where the venue publishes. The status page and the maintenance schedule, bookmarked, for every venue you actually trade on. Scheduled windows are announced; the only reason to be surprised by one is not having looked.

Understand the difference between a resting stop and a mental one. A stop order accepted by the exchange lives at the exchange and is evaluated by the exchange. A mental stop lives in a person, and a person who cannot log in has no stop at all. Binance documents its resting variants as a category — "Conditional order is a new order category in Binance Futures. It includes stop limit, stop market, and stop BBO orders" (Binance, What are Conditional Orders?) — and the same page has you set the trigger against either the mark price or the last traded price, which are not the same number in a fast market. Which type suits a given call is a separate decision, laid out in which order type turns a published entry into a position. The point here is narrower: a resting instruction survives your absence and an intention does not. Note the flip side too, since at least one venue's notice lists order cancellation and amendment among the functions that stop: an order resting before a window may be one you cannot pull during it.

Size so that an unreachable hour is survivable. This is the one that actually decides outcomes. If a position needs your attention within minutes to stay alive, the leverage on it has assumed something about connectivity that no venue promises.

Do not retry blind. When a confirmation does not come back, check the order and position screens before sending anything again, and if those screens will not load, send nothing.

Log it as downtime. Write "venue unreachable, entry missed" in your own record rather than nothing. Six months later, a folder of those entries tells you something real about your access, which is a different problem from the quality of the calls.

None of this is about whether an exchange might one day fail to give your money back — that is a separate risk with separate warning signs, treated in what happens to your money if a crypto exchange fails. This is the much more common hour in which everything is fine, your balance is intact, the venue's own notice says your assets are safe, and you simply cannot do anything.

The uncomfortable part is that the hour is not rare, and the only lever you hold during it is the one you set beforehand.