The Water Heater That Backdrafted After the Furnace Left
Why this matters
This case starts with a carbon monoxide (CO) alarm at 4:50 in the morning, 47 days after a crew replaced a gas furnace with a heat pump and drove away. Nobody was hurt. The fault was gone by the time the shop heard about it, because the gas utility, called out by the fire department, had already shut the water heater off, so the only way to find out what happened was to read the paperwork backward and then check the house against it. The paperwork turned out to be most of the answer, including the parts it did not say.
First, when a CO alarm call comes in
The office's first words do not change for a converted house: everyone out of the house now, leave the door open behind you, call 911 or the fire department from outside, and do not go back in to silence the alarm or turn anything off. Anyone with a headache, nausea, dizziness or confusion tells the 911 dispatcher so; that is a medical call, not a service call. The tech who goes later wears a personal CO monitor from the door, relights nothing the fire department or the gas utility has shut off, and leaves any red tag on until the cause is found and fixed. Everything below happened after that.
The record as the shop found it
The office pulled four documents the morning after the alarm.
SURVEY CARD Aug 12 furnace 70,000 Btu/h nat draft
WH 30,000 Btu/h nat draft
both -> exterior masonry chimney
FLAG: shared, vent check at removal
QUOTE Aug 20 remove furnace, install HP + AH
(no venting line item)
INSTALL TICKET Oct 3 furnace out, gas capped + tested
connector removed, thimble sealed
HP commissioned, customer trained
(no entry for water heater)
FIRE DEPT Nov 19 alarm peak 112 ppm, basement 60 ppm
WH shut off at valve by utility
Read top to bottom it looks like a well-run job: a survey that caught the shared chimney, a clean install, a commissioned heat pump. Read for what is missing, it is a chain with one link absent. The survey's flag never reached the quote, so no one priced or scheduled a vent check. The install ticket records the furnace side of the chimney in detail and says nothing about the appliance still using it. "Thimble sealed" does not say with what.
Two candidate stories fit the bare record. One: something new went wrong in November, such as a blocked chimney, a failed water heater, or a new exhaust fan. Two: the removal left the water heater on a vent that no longer worked for it, and it took a cold night to show. The house had to decide between them.
Checking the house against the record
The tech went back with a CO monitor, a combustion analyzer, a manometer, a mirror and a flashlight, and with the gas utility's red tag still on the water heater.
Was the chimney blocked? From the cleanout, the flue was clear: a small pile of tile fragments and dark, damp debris, no nest, no collapse. A camera up the flue showed the clay tile liner intact but wet-stained at the joints for most of its height. A blocked flue would have explained the alarm on its own, and its absence pushed toward the second story. The staining pushed further: water running down a tile liner is condensate, which forms when flue gas cools below its dew point in a flue too large and too cold for it.
Was the water heater itself bad? The burner, inspected cold, showed a light, even rust scale and no soot at the burner or the draft hood. A heater burning badly enough to make 112 ppm on its own usually leaves soot. That made a combustion fault unlikely, though only a firing test could rule it out, and the heater was tagged.
Had something new been added? The household had changed nothing: no new hood, no new dryer, no new bath fan. One thing was new since August, and it was the shop's: the furnace was gone.
What did "thimble sealed" mean? The old furnace connector opening was covered with a square of sheet metal held by foil tape. Two corners of the tape had let go. A smoke pencil at the gap was drawn in hard. That opening was feeding cold basement air into the chimney below the water heater's connector, which cools the flue and robs the remaining appliance of draft.
What did the venting tables say? The chimney is exterior, about 22 feet, with a tile liner sized for 100,000 Btu/h of combined input. The water heater alone is 30,000 Btu/h. Against the exterior masonry chimney table of the adopted fuel gas code (NFPA 54 Chapter 13 or IFGC 504), that single appliance was below the listed minimum input for that flue. The combination did not comply the moment the furnace came out. That fact was available on October 3, from a table, without firing anything.
Why the tech did not fire it to prove the story
The tempting move at this point is to pull the tag, light the heater on a cold morning and watch it spill, so the diagnosis is seen rather than inferred. The tech did not, and the reason is the rule that governs this whole case: a reproduction test on a combustion path recreates the exact condition that put 112 ppm into the house, with the household upstairs. The evidence already in hand was enough to act on. The table said the vent did not comply; the smoke pencil said the thimble was open; the wet liner said the flue had been running cold for weeks. Three independent findings pointing one way, and every one of them demands the same fix whether or not the heater spills on demand.
What firing it would have added was only confirmation, and confirmation comes cheaper after the fix: a worst-case test on the corrected vent either passes or it does not, and nobody is exposed to the old fault to get there. The tag stayed on.
Why it took 47 days
This is the part that makes this failure dangerous, and it is worth understanding rather than memorizing.
A natural-draft appliance relies on hot gas being lighter than the air around it. A tall, warm, right-sized flue makes strong draft. A large flue on an exterior wall, carrying only a small appliance, stays cold, so the first minutes of every firing push flue gas into a column of cold air that resists it. Through October that column was warm enough most of the time: mild nights, sun on the south-facing chimney through the afternoon, and short firings that spilled briefly and cleared. On the night of November 18 the outdoor low was about 15 F. The water heater had sat idle most of the night, the flue had dropped to near outdoor temperature, and the open gap at the thimble was dumping more cold air into it. At 4:30 the heater fired for an early shower, spilled at the draft hood and did not clear. The alarm, which by its UL 2034 listing integrates exposure over time, sounded twenty minutes later.
The damp liner tells the same story from the other side. Every short, cool firing for six weeks had been condensing water inside the chimney, which slowly works on tile joints and mortar. Left alone, the next failure would have been structural rather than a single morning.
The fix, and how it was confirmed
The household was offered three ways forward: a stainless liner sized from the tables for the 30,000 Btu/h heater, a sealed-combustion or power-vent water heater, or the heat pump water heater their plan had scheduled for two years out. The heater was four years old. They chose the liner, because the heat pump water heater needed a condensate route and a circuit that were not ready, and they did not want another month of cold showers.
The liner crew sized it from the adopted edition's tables, ran it full height, and capped the old thimble with a masonry seal in place of the tape. Then the tester ran the shop's removal and venting check in full, which is the procedure the October crew should have run:
- Worst-case depressurization in the basement, bath fans and dryer on: -1 Pa against outdoors, inside the protocol limit.
- Water heater fired from cold: draft established in under a minute, no spillage at 5 minutes by smoke pencil and mirror.
- Flue CO: 18 ppm air-free, under the shop's 100 ppm default.
- Re-tested under normal conditions: draft in about 30 seconds.
- CO alarms: the basement alarm that sounded and a new one outside the bedrooms, both tested.
The red tag came off only after all five, and the card went into the file.
What the shop changed
The defect was not a missing skill. The October crew could have run every one of those tests. It was a flag that lived on a document the install crew never opened. Three changes followed:
- The survey flag now writes itself into the quote. Any "shared vent" flag on a survey card adds a line item for the venting check with its labor hours, so it is priced, scheduled and visible to the crew.
- No removal closes without the venting check card. The job cannot be marked complete in the system until The Gas Appliance Removal and Venting Check SOP card is attached, signed and every field filled, including when there is no remaining appliance, which the card says explicitly.
- "Sealed" needs a noun. Every opening closed at a removal is recorded by material and method, and tape alone is not an accepted seal on a chimney opening.
What would have put this case somewhere else
An interior chimney of the same size, running warm inside the house, might have kept the heater within the table, and then the October crew's omission would have produced no symptom at all; the shop would never have learned it had the gap. That is not a reassurance. It means the same habit is sitting in other files, unexamined, and a removal audit of the past year's conversions is the cheaper way to find out. In this shop's audit, two of eleven removals with a remaining vented appliance had no test on record; both were retested and one needed a liner.
Soot at the burner or the draft hood would have sent the case the other way: a combustion fault makes CO in any vent, and the fix would have started with the burner, not the chimney. A blocked flue at the cleanout would have done the same. Each of those was looked for and was not there, which is what made the removal the explanation rather than a coincidence of timing.
References
- NFPA 54, National Fuel Gas Code, Chapter 13 venting tables, and the International Fuel Gas Code Section 504, in the edition your jurisdiction has adopted
- UL 2034, Single and Multiple Station Carbon Monoxide Alarms, for how alarm response is tied to concentration and time
- See related: What Removing a Gas Appliance Does to the Ones Left
- See related: The Gas Appliance Removal and Venting Check SOP
- See related: How to Handle the Chimney a Furnace Left Behind