This is an interesting note for those of you who might run into something similar in the future.
Most fuel pump failures are of the "sudden stop" variety. Sometimes you get odd grinding noises first, but frequently when they fail you come out to the car, put the key in, turn it and it cranks but refuses to fire because there's no fuel pressure.
This pump is out of my 2002 Suburban; it is original. The symptoms that prompted me to pull it were an apparent "out of fuel" condition at high altitude and temperature. In addition it has been "out" of fuel when the gas gauge reads that there is 1/4 of a tank remaining. Out means out too -- no start and you can hear the pump cavitate during its few-second run when the key is turned to "on". Put fuel in, it works fine.
I have detached the fuel level float arm and sender (which you can see upside down next to the pump motor); they're ok with no indication of a problem.
The interior of the tank is extremely clean; I see no residue or "trash" in it at all so the possibility of some piece of trash getting into the pickup area and blocking it is excluded. On the other hand the in-cup strainer is quite dark, implying that it is full of dirt collected over the ~20 year service life. Interestingly enough the in-tank, outside-cup strainer is not similarly clogged, which implies very strongly that whatever is in that "cup" strainer came from the cup or pump mechanism and was not in the fuel feeding the pump itself otherwise it should be equally bad in both strainers -- but isn't.
There's a one-way silicon valve at the bottom of the cup which clearly is intended to provide primary flow. It's fine. The purpose of that second, thin line is not clear; its either a feed or, quite-possibly, for pump-cooling purposes. Whatever it might be it's not cracked; I checked carefully and its hardline so vacuum capable without collapsing.
The second internal strainer picks up fuel and the pump propels it through the hose visible to the outlet. There is a mechanical pressure regulator on the fuel rail at the engine; the spill from that comes back down the other line back into the cup, so fuel that is not immediately burned is "recycled" in the cup and does not go back into the tank at-large, at least not directly. I'm not sure I like that design at all because it means if the fuel coming back from the rail is hot as it does not go into the larger volume of the tank itself, which isn't as hot. That is, the fuel in the cup could wind up having a very elevated temperature under adverse conditions in which the engine burns far less than the pump delivers but the fuel lines are hot and so is the engine. The cup is slightly less-tall than the tank profile itself; the clear intent is that when the tank is getting toward empty and fuel is sloshing around it will fill the cup to whatever level but not empty back out.
Close examination of the small line reveals no cracks or leaks from which air could be introduced, and the silicone one-way valve at the bottom looks perfectly ok. The strainer is removable; I did so and it is not plugged. Discolored yes, but not sufficiently fouled to impede fuel flow at all. The connector at the top of the physical pump motor is in good shape with no evidence of burnt contacts and the interior beyond the seal is dry.
So what the hell is going on here? Oh, and the cup had fuel in it when I pulled the pump too. Not a large amount but an approximately correct amount for the fuel I siphoned out of the tank -- it was not empty.
Specifically, that's a metal pump body so how in the hell is it drawing air when there's 10+ gallons in the tank and the cup has fuel in it? How the hell is there an air leak on the suction side of the pump? I sure can't find one on close examination but quite-clearly there is one.
So here's what I think was going on with the hot/altitude shutdowns -- the return from the fuel rail goes into the cup, not the tank at large. The valve at the bottom only admits more fuel when the cup drops below tank level. So the hot fuel that is returned gets added to a very small volume of fuel and does not dissipate much of its heat. Most of the time this is ok -- not great, but ok. Oh, and the motor's cooling is provided by this very same gas, so the harder the pump is working the more heat it is contributing to that small amount of fuel as well.
Gasoline starts to boil its lightest fractions at 95F. At 5,000' that's depressed by about 8%, so call it 90F.
Pumps don't work when attempting to pump a gas unless they're positive-displacement. So now under high altitude and hot conditions eventually the fuel in the cup gets hot enough that it boils under suction at the pump -- the pump cavitates and the fuel rail pressure collapses! You're ****ed. A key off/on allows enough time for the boiling to stop as the hot feed back to the cup and the heat addition from the pump motor stops immediately when the engine ceases running. It doesn't take much; as you know take a boiling pot off the heat and the bubbling stops almost-instantly. So when you key off/on you now suck fluid again and it starts.
The problem is that if its still smoking hot it will do it again, and maybe quite quickly. None of these trucks have fuel coolers on the return rail anywhere. I've looked on mine -- nope. All there is for "cooling" is a relatively short section of metal line tucked up under the frame where it likely gets ZERO airflow. My TDI Jetta had a fuel cooler on the return line under the pax seat where airflow under the car provides a decent amount of convection, but it's not these for gasoline -- or on these trucks. There is a short section of hard metal line but it's tucked up under the frame and body where it gets basically zero airflow (and thus won't do much of anything to cool the returning fuel.) Thus the pump can be fine but if it tries to pump vapor you get nothing.
There is no fuel pressure sensor on these vehicles so the ECU has no knowledge of the problem and thus doesn't throw a code; from the point of view of the engine computer you simply ran out of fuel.
The only way to prove this would be put a pressure gauge on the rail's test point where you can see if while driving along with power monitoring on the pump motor pins, then take the truck to altitude and attempt to provoke the fault; if you have power to the pump and no or very low and fluctuating pressure there you are. In short its pretty-much identical to vapor lock in an old non-return carb'd engine where the fuel in the shut down engine would boil in the pump at the block and prevent it from re-priming when you attempted to start it.
It seems they figured this out sometime in the last 20 years as the new pump is quite different. It has two sections in the cup. The intake side has the flapper and a suction hose to the motor compartment. The fuel return goes into this compartment as well. The second compartment contains only the motor and the output (I bet there is an intentional bleed to fill it with gas on a continual basis for cooling purposes as well.)
This is a much better design; if the returned fuel is hotter than Hell and boils it matters not, since if the level in the cup (boiling or otherwise) is lower than the tank's level fuel will be admitted from the tank into that section and, being cooler, will quench that. The suction comes from the bottom of the cup and into the drive motor, which, I presume (I didn't try powering it in a big thing of gas where I could look!) continually bleeds off a small amount of the output which will overflow back into the tank, thereby providing positive rather than passive cooling.
Betcha the kid's van has the same issue -- we shall see.


