How is this different from ordinary peeling?

By what comes away, and it is worth being precise because the two have almost nothing in common beyond the word peeling. Lift a flake and turn it over. Bare concrete, dust or a powdery grey film on the underside means the coating never bonded to the slab. That is a preparation failure, the whole system is affected, and the floor comes off. Coating on the underside means the layers failed to bond to each other. The coat below is usually still sound and still firmly stuck, which is why this one is often repairable by taking off and replacing only the layers above the failure. The two get confused because they look identical from standing height. They are separated entirely by the underside, so look before you quote.

What is a recoat window and why does missing it cause this?

It is the period during which the coat already down is cured enough to walk on and take another layer, but not yet so fully cured that the next coat has nothing to grab. Inside that window the two layers bond chemically. The surface still has reactive sites available, the new coat softens into it slightly, and what you end up with behaves much more like one thick layer than like two thin ones. Once the window closes, that chemistry is finished. The cured surface is hard, glossy and essentially inert, so the next coat can only sit on top of it and hold on mechanically. On a smooth, fully cured, unabraded surface there is very little for it to hold on to, and the result is a top coat that looks perfect for weeks and then sheets off the first time anything stresses it.

The data sheet said four hours and I recoated in three. Why did it still fail?

Almost certainly because the sheet quotes that window at a reference temperature and the job was hotter than the reference. Cure is a chemical reaction, and reaction rate climbs steeply with temperature. The usual working assumption is that the rate roughly doubles for every 10 degrees Celsius, so a slab well above the rated temperature does not shorten the window by a little, it shortens it by a lot. A window rated at two to four hours can be under two hours on a hot slab, which means a recoat at three hours was late rather than early. This is why summer work and south facing bays produce so much of this defect. The crew followed the sheet, the sheet was not wrong, and the number on it simply did not apply to the slab they were standing on. Take the slab temperature with an infrared thermometer and scale the window for it rather than reading the printed figure straight off the page.

What can sit between two coats and stop them bonding?

Anything at all, because the bond only needs interrupting rather than destroying, and most of the usual culprits are invisible. Amine blush is the most common. It is a waxy, water soluble film that forms when the curing agent reacts with moisture and carbon dioxide in the air instead of with the resin, and it can be present on a floor that looks completely normal. Wash it off with warm water before recoating, never with solvent. After that: dust settling out of the air during cure, especially anywhere grinding is still going on elsewhere in the building. Condensation, if the slab dropped near the dew point overnight between coats. Hand and foot oils from anyone kneeling on the surface. Overspray from a release agent or a lubricant used nearby. And on outdoor or partly open work, pollen and exhaust residue. The common thread is that none of these announce themselves. If a floor sat overnight between coats, assume something landed on it and prepare accordingly.

Can I recoat once the window has closed, or is it too late?

You can, but the bond has to be earned mechanically instead of chemically, and that means abrading rather than just cleaning. The method is to wash the surface first, because abrading a blushed or contaminated floor grinds the contaminant down into the profile you are about to create. Then abrade the whole area to a uniform dull finish, screen it, vacuum it thoroughly, and recoat. What you are after is that no gloss remains anywhere, because any spot still shiny is a spot the next coat is only resting on. Done properly this is a perfectly sound floor. What does not work is a light scuff of the obvious areas and a hopeful recoat, which produces a floor that fails in patches later and is far harder to diagnose than a clean failure would have been.

Does the whole floor have to come off?

Usually not, and that is the practical difference between this and a slab bond failure. The layer that let go is the one that has to come off, along with everything above it. The coat underneath is generally still sound and still bonded, so it can stay and become the surface you prepare and recoat onto. That is a much smaller job than removing a full system back to concrete. Test the boundary before you commit to it. Grind back into the sound area, feather the edge, and try to lift the surrounding coating with a blade. Material that resists is telling you the failure is confined. Material that keeps releasing as you work is telling you the whole intercoat bond is marginal, and then the honest answer is that everything above the sound layer goes.

Is it always the installer at fault?

More often than with most defects, because nearly every cause is a decision made between two coats. The recoat window, the surface preparation between layers, whether blush was washed off, whether the floor was protected from dust overnight and whether the products were meant to go over each other are all choices. So is checking the slab temperature rather than the printed figure. The cases that are genuinely not the installer are narrower: a product data sheet that states a window without stating the temperature it applies at, two products from different manufacturers sold as compatible that are not, and a building where somebody else generated dust or ran heating during cure. Those happen, but they are the minority. What settles it in practice is the record. Times, slab temperatures and what was done to the surface between coats, written down at the time, tend to name the cause without much argument.

Intercoat Delamination: When One Coat Lets Go of Another

The underside of the flake settles this in about two seconds. Concrete on the back means the bond to the slab failed and you are on the wrong page. Coating on the back means the layers never bonded to each other, and that is almost always a recoat window that closed before anybody expected it to, or something invisible sitting on the surface when the next coat went down.

Quick answer

Intercoat delamination is a top coat releasing from the coat below rather than from the slab, and the underside of one lifted flake settles it. Coating on the back means the layers never bonded to each other. Concrete on the back means something else entirely and you want the peeling page. Almost all of it comes down to two things: a recoat window that closed earlier than anybody expected, or something invisible sitting on the surface when the next coat went down.

The flake test, and why it is the whole diagnosis

This family has three members and they look the same from standing height. One observation separates all three, and it takes a blade and about two seconds.

Coating on the underside

One coat let go of another. This page.
The layer still on the floor is usually sound, glossy and firmly stuck
Flakes tend to come away in wide thin sheets that hold their shape
Scope: the failed layer and everything above it. The coat below can often stay

Bare concrete or dust

The bond to the slab never formed, which puts the cause in the preparation
The exposed slab may look glassy, polished or powdery rather than profiled
Scope: the whole system, because the same preparation is under all of it
Go to peeling

Only where a vehicle parks

Tire shaped patches, with the missing coating usually stuck to the tread
Walkways and the middle of the bay are untouched, which is the giveaway
Scope: depends on why it let go, and the product class decides most of it
Go to hot tire pickup

Cause one: the recoat window closed

A recoat window is the period during which the coat already down is cured enough to take another layer but not yet so fully cured that the next one has nothing to grab. Inside it the two layers bond chemically: reactive sites are still available, the new coat softens into the old one slightly, and what you get behaves much more like a single thick layer than like two thin ones.

Once it closes, that chemistry is over. The cured surface is hard, glossy and effectively inert, so the next coat can only rest on top and hold on mechanically. On a smooth unabraded surface there is almost nothing to hold, which is why this failure so often looks perfect for weeks and then sheets off the first time the floor is stressed.

The window on the sheet is not the window on your slab

Recoat windows are quoted at a reference temperature, and cure rate climbs steeply as it gets hotter. Enter the slab temperature you actually have and the tool scales the window, so you can see how much of it you really had rather than how much the sheet promised.

tradeepoxy.com/calculators/polyaspartic-cure-time-calculatorStatic preview

A window rated 2 to 4 hours at 70F is under 2 hours once the slab reaches 90F

Open the live calculator →

On the medium preset that window is rated 2 to 4 hours at the reference temperature of 70F. The same window runs 0.9 to 1.9 hours on a 90F slab, because cure rate roughly doubles for every 10 degrees Celsius. That is not a small correction, and it is the arithmetic behind most summer intercoat failures: the crew read the sheet, followed it, and recoated at three hours into a window that had already shut.

Note which direction this runs, because it is the opposite of the mistake people expect. A cold slab stretches every phase and mostly costs you time, which is the case the hot tire pickup page argues from. A hot slab shortens the window and costs you the bond. Take the slab temperature with an infrared thermometer rather than reading the room thermostat, and take it in the bay that gets the sun rather than the one by the door.

Cause two: something was on the surface

The bond only needs interrupting, not destroying, and most of what interrupts it is invisible on a floor that looks perfectly clean.

Amine blush, by a wide margin the most common. A waxy, water soluble film left when the curing agent reacts with moisture and carbon dioxide in the air rather than with the resin. It can sit on a floor that looks completely normal, and it is why blushing matters far more than its cosmetic appearance suggests. Warm water and a mild detergent, never solvent
Condensation, where the slab dropped near the dew point overnight between coats. A film of water is a perfect release agent and it is gone by the time anyone looks
Dust, particularly anywhere grinding is still going on elsewhere in the building. Airborne dust settles onto a curing floor for hours after the work that made it stopped
Hand and foot oils from anyone kneeling on the surface to work, and overspray from any release agent, lubricant or tire dressing used nearby

If a floor sat overnight between coats, assume something landed on it. That single assumption prevents most of this defect, and acting on it costs a wash and a screen.

Cause three: the two products were never going to bond

Less common than the first two but worth ruling out, because no amount of preparation fixes it. A top coat has to be formulated to go over the base it is being put on. Mixing manufacturers, putting a rigid coat over a flexible one, or coating over a product that is still releasing solvent all produce failures that look exactly like a missed window.

The tell is that it fails uniformly and early across the entire floor rather than in the areas you would expect from a timing or contamination problem. Check both data sheets for a stated compatible system before assuming anybody mistimed anything.

Recoating once the window has closed

This is entirely doable. The bond simply has to be earned mechanically instead of chemically, which means abrading rather than cleaning, and the order matters more than people expect.

Wash first. Abrading a blushed or contaminated floor grinds the contaminant down into the profile you are about to open, which is the opposite of what that profile is for
Abrade the whole area, not the obvious patches, to a uniform dull finish. Any spot still showing gloss is a spot the next coat is only resting on
Vacuum thoroughly, then tack the surface. The abrasion you just did is now the dust that ruins it
Recoat promptly, because the clean abraded surface starts collecting the same contaminants that caused the original failure

Done properly the result is a sound floor. What does not work is a light scuff of the visibly loose areas and a hopeful recoat, which produces a floor that fails in scattered patches later and is much harder to diagnose than a clean failure would have been.

What can be saved

More than with a slab bond failure, and that is the practical consequence of reading the underside. The layer that let go comes off along with everything above it, but the coat underneath is generally still sound and still bonded, so it can stay and become the surface you prepare onto.

Test the boundary before committing to a scope. Grind back into the sound area, feather the edge, and try to lift the surrounding coating with a blade. Material that resists is telling you the failure is confined and a sectional repair is honest. Material that keeps releasing as you work is telling you the intercoat bond is marginal across the floor, and then everything above the sound layer goes.

Preventing it on the next floor

Scale the recoat window for the slab you have. The printed figure is quoted at a reference temperature, and a hot slab can more than halve it
Write the times down as you go. When each coat went down, the slab temperature and the humidity. This defect is argued about after the fact more than most, and a record ends the argument
Wash between coats if there was any gap at all, and treat blush as present rather than as something you would have noticed
Protect the floor overnight. Stop other trades generating dust in the space, and check the slab against the dew point before the next coat if it sat through a cold night
Keep the system to one manufacturer unless you have a written statement that the two products are a compatible system