I have stopped many times and checked the fuel pressure (always around 40psi), checked for bubbles (none noticeable), checked for water or crap in the fuel - from the check valve (none noticeable).
I have taken the old (replaced) pumps and bench tested them and have found no problems and they have good pressure
When you pull over and check the fuel pressure, the engine is idling and under no load so the injector pulse width is minimal = great restriction on the pump. Under those conditions, it is easy for the pump to maintain pressure. I would be curious to know what happens to the rail pressure when you drive off, the pulse width widens, the injectors open more and therefore the restriction goes way down. Can the pump maintain the same pressure as a new one under those conditions? I would rig up a gauge I could see while driving or design a bench test with a variable restriction I could adjust and compare new vs suspect pumps.
It would also be good to know, based on the tank design, where the various components are situated relative to the gas level when the problem begins to occur. For example, is the pump body exposed but the pickup tube completely immersed and so on. This may be a clue as to what could be degrading over time. Pinholes, leaks etc. These pumps are good at making pressure at the outlet but their design makes them poor suction devices. The inlet conditions are important and this design could be very susceptible to changes or variations in operating conditions. Just some thoughts...