A hotel pool’s safety system spotted the problem and sounded the alarm. Nobody heard it, because the alarm was an email and the email never arrived. That is roughly what Phantasialand told German broadcaster WDR after 15 guests at its Hotel Ling Bao were hurt by too much chlorine in the pool during last week’s Microsoft 365 outage.
Key takeaways
- Phantasialand told WDR that a pool chlorine alarm went out by email and nobody saw it because of the Microsoft outage.
- Email retries and queues messages, which is fine for invoices but too slow and too silent for safety alarms.
- Give every critical alert a second, independent channel, such as a push notification, SMS or phone call from another provider.
- Add a heartbeat check so that silence from a device or script raises the alarm too.
If anything you run sends its warnings by email alone, this guide is for you. You’ll see what happened, why email is a weak alarm channel, and five practical steps to make sure the next critical alert reaches a human even when your mail provider is down.
What happened at Phantasialand
According to WDR, the automatic dosing system in the hotel’s sauna and pool area pumped too much chlorine into the pool. The maintenance safety system noticed and raised the alarm by email. Phantasialand told WDR that because of the Microsoft outage, no mails were coming in, so nobody saw the warning.

Guests who had used the pool complained of health problems, and the fire brigade and paramedics were called. Radio Erft reports that crews from Brühl and the neighboring towns of Erftstadt, Hürth, Kerpen and Wesseling checked 15 people. According to the fire service report it cites, 14 had light injuries and one a moderate injury. Four were taken to hospital. WDR says early reports of five hospital patients were corrected to four, and that one person stayed in hospital for observation. Tests on site found a defect in the dosing system, the water was replaced, and the fire service told WDR on Saturday that the pool had reopened.
Two things are worth keeping apart. The chlorine came from a faulty dosing system, not from Microsoft. The outage is what stopped the warning from getting through, according to the park’s own account. Local outlets also differ on which day the incident happened, so treat the exact timing as unsettled.
By the numbers
15people checked by paramedics
4taken to hospital
100,000outage reports within an hour
The outage behind the missing emails
The wider outage started on Thursday, 12 November 2026, and dragged into Friday. ABC News reports that Downdetector logged over 100,000 reports within an hour of Microsoft’s first acknowledgment. On Friday morning Microsoft warned of “intermittent access issues or delayed email delivery” while queues cleared.
Microsoft says it wasn’t an attack. In a statement quoted by ABC News, the company said “the incident was introduced by an issue during an earlier network capacity upgrade that surfaced following a recent network change.” It has not yet published a full post-incident report. Our Microsoft 365 outage explainer covers the timeline, who was affected and why old emails kept turning up days later.
Why email makes a fragile alarm
Email was built to get there eventually, not to get there now. When a server can’t deliver a message, the sender usually queues it and retries. That’s great for invoices, but a chlorine warning that turns up six hours late is no use to anyone.
A few other weak spots:
- One provider, one point of failure. If your alarm sends to a mailbox hosted by the same company that is down, the warning has nowhere to go.
- Silent failure. A sending device usually can’t tell that nobody read the message. It sent it, so as far as it knows the job is done.
- Noise. Alerts that land in a busy inbox compete with newsletters. Google’s SRE book warns in its monitoring chapter that when pages come too often, people “second-guess, skim, or even ignore incoming alerts.”
The same chapter says the parts of a monitoring system that page a human “need to be very simple and robust.” An alarm that silently depends on one cloud mailbox is neither.
Did you know?
A device that sends an alarm email usually can’t tell whether anyone read it. As far as it knows, the job is done once the message leaves.
How to make critical alerts survive an outage
These steps work for a small business, a home lab or a building system that can call a web address. You won’t need all five, but critical alarms should get at least the first two.
Step 1: List every alert that matters
Write down every system that warns you about something dangerous or expensive: pool or boiler controllers, freezer and server-room temperature sensors, backups, fire panels, payment failures. For each one, note where the alert goes today and which company carries it. If the answer for a safety alarm is “an email to one shared inbox,” it goes to the top of the list.
Step 2: Add a second, independent channel
The fix is redundancy, and it is the advice monitoring vendors give themselves. Healthchecks.io’s notification docs recommend setting up “two different notification channels (for example, email and Slack)” so that if one transport fails, “you still receive a notification over the other channel.” For vital services they suggest chat, email and SMS together.
Independent means a different company and a different path. A Teams message is no backup for an Outlook email, because both ran on Microsoft 365 and both were hit last week. A text message, a phone call or a push notification from another provider is.
Step 3: Send a push alert with one command
If your device or script can make an HTTP request, ntfy is a quick way to get alerts onto a phone. You install the ntfy app, subscribe to a topic, and anything posted to that topic pops up as a notification. There’s no sign-up, so the docs warn that “the topic is essentially a password”. Pick a long name nobody will guess.
curl \
-H "Title: Pool chlorine high" \
-H "Priority: urgent" \
-H "Tags: warning" \
-d "Dosing alarm in the pool area. Check now." \
ntfy.sh/example-pool-alerts-7f3k9q
The Priority header sets how loudly the phone notifies you. ntfy also accepts X-Priority with a number, where 5 is the most urgent. On Android you can give each priority its own sound and vibration pattern, so urgent alarms are hard to sleep through.
ntfy is open source and can be self-hosted. Our guide to turning an old Android phone into a home server shows one cheap place to run small tools like this.
Step 4: Add a heartbeat so silence raises the alarm
Steps 2 and 3 help when an alert is sent but one channel is down. They don’t help when the sending device itself dies, loses power or loses its network. For that you need a dead man’s switch.
The idea is simple: the system checks in on a schedule, and you’re alerted when it stops. Healthchecks.io’s cron guide describes it as heartbeat monitoring. Your job sends an HTTP “ping” each time it runs, and “when Healthchecks.io does not receive the HTTP request at the expected time, it notifies you.” Its example adds a curl call after a daily backup:
8 6 * * * /home/me/backup.sh && curl -fsS -m 10 --retry 5 -o /dev/null https://hc-ping.com/your-uuid-here
The extra options retry failed requests, cap the run time at 10 seconds and keep it quiet unless there’s an error. Point the heartbeat’s own alerts at your second channel, not at the mailbox you’re trying to protect.
Step 5: Get told when your provider is down
If your organization runs on Microsoft 365, an admin can watch the Service health page under Health in the Microsoft 365 admin center. Microsoft also offers email notifications for new incidents, but during a mail outage those emails may be stuck too.
Microsoft’s own page points to two backups: the public service status page when you can’t sign in to the admin center, and the @MSFT365status account on X. When your provider is down, treat that as an alarm in its own right. Check your critical systems by hand until mail flows again.

Common mistakes and fixes
- The backup channel runs through the same provider. Fix: use a different company, such as SMS or a push service, for at least one route.
- Nobody tests the alarm. Fix: send a test alert through every channel on a schedule, and make sure a real person confirms it arrived.
- Alerts go to one person. Fix: send critical ones to a rota or group, so a phone on silent doesn’t end the chain.
- Too many alerts. Fix: page people only for urgent things they can act on, and send the rest to a daily summary. Noisy alarms train people to ignore them.
- An easy-to-guess ntfy topic. Fix: use a long random topic name, or a self-hosted server with access control.

What it means for you
Most of us will never manage a hotel pool, but almost everyone depends on an email that “should” arrive: a backup report, a server warning, a smoke-alarm notice from a smart home hub. Last week showed how fragile that is.
Spend an hour this week on Step 1. For anything that could hurt someone or cost real money, add a second channel and a heartbeat. If an outage like this hits again, you’ll find out from your phone instead of from the fire brigade.
For more on reacting when a service you rely on goes dark, see our guide to checking whether Claude is down and our explainer on the Workday outage. More guides are in Software & How-To.
Frequently asked questions
Did the Microsoft outage cause the Phantasialand chlorine accident?
No. Tests found a defect in the pool’s automatic dosing system. According to what Phantasialand told WDR, the Microsoft outage stopped the safety system’s alarm emails from arriving, so nobody saw the warning in time.
What is a dead man’s switch in monitoring?
It is a check that expects a regular signal, or heartbeat, from a device or job. If the signal stops arriving on schedule, the monitoring service alerts you, so a dead device or broken script cannot fail silently.
Is Teams a good backup if Outlook goes down?
Not for critical alerts. Teams and Outlook both run on Microsoft 365, and both were hit in the November 2026 outage. A backup channel should come from a different company and travel a different path.
How do I check if Microsoft 365 is down?
Admins can open Health then Service health in the Microsoft 365 admin center. If you can’t sign in, use Microsoft’s public service status page or the @MSFT365status account on X.
Quick quiz
What does Healthchecks.io do when a heartbeat ping doesn’t arrive on time?
- Restarts the server
- Notifies you
- Deletes the check
It stays silent while pings arrive on time and alerts you when one is late.
Why should an ntfy topic name be long and random?
- There is no sign-up, so the topic works like a password
- Short names cost money
- Phones reject short names
ntfy has no sign-up, so its docs warn that the topic is essentially a password.
Which is a truly independent backup for an Outlook alarm email?
- A Teams message
- A second Outlook mailbox
- An SMS or push alert from another provider
A backup has to use a different company and path; Teams and Outlook both failed in the outage.
Sources
- WDR: IT-Panne Schuld für Chlorgasunfall im Phantasialand
- Radio Erft: Phantasialand: 15 Verletzte nach Defekt in Hotel-Pool
- ABC News: Microsoft apologizes for email disruption, says there was no hack
- Microsoft Learn: How to check Microsoft 365 service health
- Healthchecks.io: Configuring notifications
- Healthchecks.io: How to monitor cron jobs
- ntfy docs: Sending messages
- Google SRE book: Monitoring Distributed Systems





Write a response