April 2026 — on the gap between a well-researched plan and a wet crawl space

The design article made it sound rational. The build article made it sound competent. This one is the honest one.

Fifteen years on, here is what actually happened.

First, a correction

The previous articles open with “replacing a gas boiler.” This is wrong. We had city heat — stadsverwarming — a perfectly functional district heating connection that works reliably and requires zero maintenance.

The actual motivation was different: could you build a sustainable heating system that works economically for a normal person? Not a wealthy enthusiast with a contractor budget and a detached house with a south-facing roof. A regular apartment, tools you own, skills you have, help from people you know. Could the numbers actually add up?

This distinction matters, because it changes what counts as success. The system didn’t need to be perfect. It needed to be better than the alternative — financially, practically, and educationally — for someone who built it with their own hands rather than paying someone to install it.

By that measure, it was a success. A complicated, frequently wet, occasionally expensive, permanently educational success.

The buffer tank that buffered nothing

The centrepiece of the design was a 2,300-litre underground thermal buffer. Excavated by hand. Reinforced concrete walls, rebar, wooden formwork. A hoist to lower materials in. It took months. It is an impressive hole.

It has never been used.

The reason is something no forum post mentioned: three concrete floors of an Amsterdam apartment building contain an enormous amount of thermal mass. The building is the buffer. Heat the floors at night, the building holds it. The floor heating was already doing what the underground tank was designed to do, for free, using infrastructure that already existed.

The 2,300-litre tank is still there. It is very well insulated. It is empty. You can fit a surprising number of bicycles in a well-insulated underground concrete box.

Dutch weather: a solar thermal system’s natural enemy

The design assumed a climate that produces useful solar thermal energy in winter and manageable solar thermal energy in summer. This describes many places. It does not describe Amsterdam.

Dutch winters are grey. Not occasionally grey — structurally, persistently, almost philosophically grey. The sky commits to overcast in October and reconsiders in March. A vacuum tube solar collector in January in Amsterdam is a beautifully engineered device for collecting the thermal energy of a ceiling. On the rare cold sunny days when the system would work well, it does. There are perhaps twelve of those days between October and April.

Dutch summers are the opposite problem. The collector produces heat at a rate the system cannot consume. The buffer fills. The floor heating is off. There is nowhere for the energy to go. The fluid in the solar loop starts to boil.

Boiling fluid in a closed system does unpleasant things. The boiling causes a redox reaction in the copper pipework at the hottest point — the connections near the manifold. The copper oxidises. The fittings corrode. The joints fail.

I replaced those pipe sections multiple times. Not because the soldering was bad. Because the physics was non-negotiable: too much heat, no load to absorb it, and copper that has opinions about that situation.

The collector I machined myself

The warm water collector was a custom build — copper tube, brass fittings turned on a lathe in my father’s workshop. Precision work. Every fitting machined to exact dimensions, assembled carefully, cold-soldered at the joints, reinforced with epoxy and fibreglass because belt-and-suspenders seemed wise.

The cold-soldered joints lasted several years. Then I came back from Spain to a wet wall.

Thermal cycling — expansion and contraction, summer to winter, hot to cold, thousands of cycles over several years — had worked the joints apart. The epoxy and fibreglass held the fitting in place. The fitting itself was no longer sealed. The wall had been quietly absorbing water for however long it took for “quietly” to become “visibly”.

The lesson: in a solar thermal system, “reinforced with epoxy and fibreglass” is not a substitute for a properly soldered joint. The reinforcement holds the shape. It does not hold the seal. What moves will keep moving until something gives, and eventually something gives at 3am on a Thursday while you are in another country.

The hole in the concrete wall

Getting pipes through a 23-centimetre reinforced concrete wall requires a core drill. I did not have a core drill. I had a hammer drill, a collection of masonry bits, and a geometric insight: a circle can be approximated by enough smaller holes arranged in a circle, and if the holes overlap sufficiently, what remains in the middle can be persuaded to exit.

I drilled a ring of 22mm holes. Then I convinced the remaining concrete to become rubble. Then I cleaned up the resulting aperture into something approximately round.

It worked. It is not the method a construction company would use. A construction company would also charge more for the job than the entire heating system cost to build. When your budget is finite and your tools are what you have, you work with what you have.

The epoxy t-shirt

The flue penetration through the roof needed a weatherproof collar — a shaped boot that would seal the gap between the stainless flue and the EPDM membrane while allowing thermal movement.

The correct solution is a custom lead flashing or a proprietary flue boot. The practical solution, when you are standing on a flat roof in November with materials you actually have, is a t-shirt.

Specifically: a t-shirt used as a mould, soaked in epoxy, formed around the flue and the roof membrane interface, allowed to cure, and then detailed with additional sealant at the edges.

That collar is still there. It has been through fifteen Dutch winters and fifteen Dutch summers. It has never leaked.

The cold-soldered precision-machined fittings: failed. The epoxy t-shirt: fine.

Engineering is humbling.

Wood: fine until the geopolitics changed

The wood stove worked well. Genuinely well. The back boiler fed the buffer, the floor heating ran, the apartment was warm in a way that only a 22°C floor can make it warm.

Then Russia invaded Ukraine, European energy markets went sideways, and the price of firewood tripled.

The wood stove had been chosen partly because wood is cheap, locally available, and price-stable. Two of those three things stopped being true in 2022. A system designed around cheap wood is a system that is now more expensive to run than city heat, which was the thing we were trying to improve on.

You cannot design around a land war in Europe. I am not sure what the lesson is here, except that “stable input prices” is not a property you can engineer into a fuel supply.

The fountain

Last autumn I heard something from the crawl space. A sound that took a moment to identify, because it is not a sound you expect to hear from a crawl space. A fountain.

Here is what had happened: the previous summer, the overheating mechanism had activated — correctly, exactly as designed — to dump system pressure when the solar loop boiled. The pressure relief valve opened. The pressure dropped. The makeup water valve, which opens automatically when system pressure falls below 1.5 bar, also opened — correctly, exactly as designed — and began refilling the system from the mains.

Except the pressure relief valve had not returned to its fully seated position after the overheat event. So the system pressure never stabilised. The makeup valve kept seeing low pressure. The makeup valve kept opening. The water went somewhere. The somewhere was the crawl space.

I broke open the wall. I found the problem. I fixed the valve. I dried out the crawl space.

A fifteen-year-old system in which two valves interact in an unexpected failure mode during an extreme summer event is not a poorly designed system. It is a system that has been running for fifteen years. Everything fails eventually. The question is whether you understand it well enough to fix it yourself, which I did, because I built it.

What it was actually worth

The system did not perform as modelled. The buffer tank was never used. The solar thermal underdelivered in winter and overcorrected in summer. The wood stove became expensive. Various joints, pipes, and valves failed in sequence over the years. I came home to a wet wall once and a flooded crawl space once.

And I would do it again.

Not because the engineering outcomes were good — they were mixed. Because living with a system you built is categorically different from reading about one on a forum. You learn what thermal stratification actually sounds like when it works. You learn what a glycol loop smells like when it gets too hot. You learn that a 23-centimetre concrete wall is passable with patience and geometry, that an epoxy t-shirt outlasts precision fittings, and that the building itself often does the job you spent months engineering a solution for.

Forum posts give you information. Building something gives you understanding. The difference between them is exactly as large as a 2,300-litre underground tank that is now storing bicycles.