Blame the snowman ...
or:
Why I fly Warriors, in nice warm weather.
Have you noticed how dirty a snowman gets as it melts?
(The situation on the approach to LHR was worse than worst case design spec, so look at worst case design spec. Any numbers are very approximate, but spurious precision has been retained so that their origin is recognisable)
900Kg of 'unusable' fuel in the centre tank - recovered to the outboard (high) end of the main tanks by the fuel scavenge system when main tank pumps are working. That's 2000 lbs or 250 UK gallons.
Up to 138 (US?) gallons of water trigger the water-in-fuel warning - 115 UK gallons. So the centre tank dregs to be scavenged could be half water, half fuel. It gets scavenged 'water first' in the cruise.
The water (mostly) comes in with ambient air replacing fuel used. At altitude the air is cold and even if saturated has relatively low water content. In cloud, the air is saturated and carries suspended droplets as well. Climbing uses fuel at the greatest rate. Climbing on the centre tank through cloud brings water into the centre tank at the greatest rate - but it is a relatively short phase of a long haul flight.
The water droplets, either carried in as cloud or condensing with adiabatic expansion of the air in the tanks, collect in the fuel, and slowly settle at the water scavenge pump inlet (at cruise attitude) and get 'burned off' when the scavenge pump discharges them adjacent to the inlet of the pump supplying the engine. Water only accumulates at the water scavenge inlet if it arrives faster than the scavenge pump sucks it away. Nothing can go wrong ...
But what if it is not water? What if the local condensation and ingested cloud is ice?
Ice granules will not coalesce to form droplets, so the layer of ice granules at the bottom of the centre tank will only be scavenged near the scavenge pump inlet. Instead of flowing to the lowest point as liquid water would, the granules will roll down a local embankment of granules, under the gravitational influence of their small density difference. So whilst the scavenge pump will keep the local area clear, ice crystals will settle like snow everywhere else. The scavenge pump will only start to clean the whole tank when the tank temperature rises above freezing, and the crystals melt into droplets and globules of water that run down to the scavenge point. If the whole take-off, climb and cruise has been in sub-zero temperatures, the centre tank water scavenge pump will be off before this happens. Centre tank clearance of water will start when (and if) the centre tank temperature rises above freezing in the descent, when it will be cleared to the main tank by the fuel scavenge pump, and burned off from the main tank by the main tank water scavenge.
If the centre tank does not collect enough heat to provide the latent heat needed to melt the ice slush before engine shutdown, the slush will melt on the ramp and collect in the sump, potentially causing a water warning at next start-up. This will clear when the water scavenge pump burns it off. If sub-zero ramp and take-off temperatures freeze this water, it almost certainly stays frozen until the next descent. There is no scavenge whilst burning fuel from the centre tank because the inlet is frozen solid. There is then a double dose of solid and powdered ice to be cleared by the fuel scavenge pump when it melts during the next descent.
Provided there is a big enough safety gap between bottom level of useable fuel and the top level of water, and assuming the scavenge pumps can't pump water quickly enough to bring the aircraft down, nothing can go wrong ....
But the snowman effect spoils the party. As a snowman thaws, the atmospheric dirt and dust collected by the falling snow (and, admittedly, dirt collected by the youngsters who built the snowman) is caught and concentrated by its receding and shrinking surface. This is not just because the dirt cannot evaporate. Even a melting snowman seems able to cling on to his dark coat whilst shedding melt water. The dirt is efficiently concentrated on the surface. So after several journeys in freezing or near freezing ground temperatures, or maybe a couple of such journeys through a particularly dirty industrial atmosphere, it is easy to visualise the dregs of the centre tank water and slush, with all its accumulated micro-muck, being slopped into the fuel scavenge area by a change of trim or even by the unstable progression of the melt process.
So water with concentrated micro-dirt surges into both fuel scavenge areas of the centre tank, to be promptly fed to the main tanks, draining 'fairly promptly' to the water scavenge points on the main tanks, to be then fed to the engines - affecting both engines within a few seconds of one another.
The dirt nucleates cavitation, killing throughput of the LP engine pump.
Fire away ...