The Water Shutoff Closed at 3am

Why this matters

An automatic water shutoff that closes in the night is doing one of two things: saving a house, or teaching its owner to switch it off. From a phone at 3am the two look identical, and the instinct is to reopen the valve so the alarm stops. This case walks the call from the alert to the cause, working through four candidates on evidence from the valve's own flow log, and includes the one reading that pointed confidently at the wrong answer.

First, treat it as a real leak

Until the evidence says otherwise, a closed valve means water may have escaped somewhere in the house. The sequence on any after-hours closure, before diagnosis:

• Nobody reopens the valve remotely. Reopening on a real leak restarts the damage at full line pressure, with nobody there to see it.
• Whoever enters the house checks before stepping in. Standing water on a floor with receptacles, a panel, a water heater or a furnace near it may be energized. Do not walk into it; leave and call an electrician or the utility to cut power before anyone enters. Do not stand under a ceiling that is bulging or stained and wet, because saturated drywall comes down without warning.
• If the house has fire sprinklers fed from the domestic main, confirm the valve sits downstream of the sprinkler takeoff, as it should. If it does not, the house has no sprinkler protection while the valve is closed, and that is now urgent information for the owner.

The call

The customers were on their first trip away since the shop installed a flow-based shutoff with eight spot sensors, three weeks earlier. The system was in away mode, which the installer had set to close after 60 minutes of continuous flow. At 3:00 on a Sunday morning the valve closed and both owners' phones got the alert. They called the shop's after-hours line.

The on-call tech asked for two things before anything else: a screenshot of the flow log for the night, and whether any spot sensor had alarmed. No sensor had. The log came through a few minutes later.

FLOW LOG  Sat night  (AWAY: close at 60 min continuous)
01:58  0.0 gpm   idle
02:00  1.9 gpm   flow start
02:11  0.4 gpm
02:46  1.6 gpm
02:57  0.8 gpm
03:00  0.8 gpm   60 MIN CONTINUOUS - VALVE CLOSED
03:00  house pressure 58 psi
03:04  house pressure 21 psi
03:08  house pressure  4 psi

The valve did what it was set to do: flow started at 02:00 and was still running at 03:00, which is exactly 60 minutes. The question was what drew water steadily for an hour starting on the stroke of two, at a rate that stepped up and down.

Candidate one: irrigation

The house's irrigation came off the main downstream of the valve, by the customer's choice at install, so irrigation flow counts. It was the first suspect because it runs at night.

It did not fit on two counts. Rate: the install record noted the irrigation zones each drew roughly 8 to 10 gallons a minute when tested, against 1.9 at its highest on this log. And time: the customers read their irrigation controller's history from its app, and the first zone was scheduled for 04:00, an hour after the valve closed. Irrigation was out.

Candidate two: a running toilet

A toilet whose fill valve sticks open, or whose flapper leaks badly enough to keep the fill running, draws water continuously and produces exactly the kind of all-night flow a duration rule exists to catch.

Two things argued against it. A stuck fill valve flows at a steady rate set by the valve and line pressure; it does not drop to 0.4, climb back to 1.6 and settle at 0.8. And a toilet fault does not start on the exact minute of two o'clock. A leaking flapper that drains slowly and refills in bursts does produce a stepped profile, but its bursts recur every few minutes or tens of minutes with idle gaps between them, and this log never returned to zero for an hour. The toilet was set aside for the neighbor to check by eye, not eliminated outright.

Candidate three: a real leak, and the reading that said yes

This is where the case nearly went wrong. The log's last three lines showed house pressure falling from 58 psi to 4 psi in eight minutes after the valve closed. With the valve shut, nothing should draw water, and a plumbing system that cannot hold pressure has a leak somewhere downstream. That is precisely the test many systems run on purpose to find hidden leaks.

The tech's first read was the obvious one: pressure decay after closure, a leak somewhere in the house, keep the valve closed and get someone inside. That instruction stood. But the pressure reading only proves that something downstream is releasing water. It does not say whether the something is a broken pipe or an appliance doing its job. A leak and a device that is mid-cycle and discharging to a drain look the same on a pressure gauge.

So the pressure reading was evidence that water was leaving the system, which the flow reading already showed. It was not independent evidence of a leak.

Candidate four: the water softener

The house had a water softener, installed years before the shutoff. A softener periodically regenerates: it backwashes its resin bed, draws brine through it, rinses it and refills its brine tank, sending most of that water to a drain. Many softeners regenerate at a set time in the small hours so nobody draws hard water during the cycle, and 02:00 is a common factory default.

The stepped profile fits a regeneration well. The phases run at different rates: a backwash at a higher flow, a slow brine draw, a rinse, a refill. The rates on this log are illustrative of that pattern rather than a match to any model's published figures, which live in the softener's manual. And a softener mid-cycle when its supply is cut keeps trying to send water to drain, which is exactly what bleeds house pressure from 58 psi to 4 psi.

The neighbor's walk, and what it settled

A neighbor with a key went in at about 3:40, told first to look at the floor before stepping into each room near equipment. What they reported, in order:

• No water on any floor, under any sink or around the water heater, and every spot sensor dry, as the app had already said.
• The toilets silent. No hiss of a fill valve, no water moving in the bowls. That settled the toilet candidate.
• The softener's display showing a regeneration in progress, and a gurgling at the floor drain where the softener's drain line ended, which faded as the remaining pressure ran out.

That closed it. Irrigation was out on rate and time, the toilets were silent, there was no water on any floor, and the softener was visibly in a regeneration cycle that started at 02:00.

Why it happened on this night and not the others

The softener had regenerated every few nights since the shutoff went in, without closing the valve. The install record explained the difference. At home, the valve's continuous-flow limit was 2 hours; the regeneration, which ran a little over an hour, fit inside it. Away mode used 60 minutes. This was the first night the house had been in away mode while the softener regenerated, so it was the first night the regeneration was longer than the limit.

The softener's own settings showed why it regenerated at all with nobody home. It was set to regenerate on a timer, every set number of days regardless of use, rather than on demand, after a metered volume of water has passed through it. A demand-initiated softener in an empty house uses no water and so has no reason to regenerate.

Choosing the fix

Three options, each with a cost:

• Raise the away-mode limit to 90 minutes. It clears the regeneration with some margin. It also lets a real leak at the same rate run longer before closing: at 1.9 gallons a minute, 60 minutes is 114 gallons and 90 minutes is 171 gallons, half again as much water on the floor in a house nobody is watching.
• Move the softener's regeneration time and leave the limit. It changes when the trip happens, not whether.
• Switch the softener to demand-initiated regeneration. The softener's control supported it; the original installer had simply left it on the timer. In an empty house it then regenerates only after real use, and in a lived-in house it regenerates when the resin actually needs it.

The shop recommended the third, kept away mode at 60 minutes, and confirmed with the customers by phone before changing anything. The softener's control was changed on the next visit, by the shop that services it.

Reopening came first, that night. With the cause established, the tech had the neighbor open the valve from the customer's app, then watch the floor drain and the softener for several minutes while the regeneration finished on restored pressure. Flow resumed, the cycle completed, the log returned to zero, and pressure recovered to 58 psi and held.

How they confirmed it held

At the learning-period return a week later, the tech read the full flow history. The softener on demand had regenerated twice in eight days of normal household use, both times while the system was in home mode, each run a little over an hour, and neither closed the valve. The customers' next trip, five nights away about a month later, logged no regeneration and no closures. The tech also wrote into the record that the softener's regeneration runs longer than the away limit, so the next person to touch either device knows the two are linked.

What to take to the next 3am call

Ask for the flow log first, and read three things off it before forming a view: when the flow started, whether its rate was steady or stepped, and whether it ever returned to zero. Then remember that pressure falling after a closure tells you water is leaving, not why. Every appliance that runs a cycle to drain (a softener, a reverse-osmosis unit flushing, a boiler auto-fill, a humidifier) will bleed pressure the same way a broken pipe does.

References

• Manufacturer manuals for the shutoff valve (flow rules, modes, pressure logging) and the water softener (regeneration modes and cycle times)
• See related: Leak Sensors and Automatic Water Shutoffs
• See related: Installing and Testing an Automatic Water Shutoff
• See related: Walking a Customer Through a Reset Over the Phone