Executive  Summary: We DID find a small potential leak from the Right ambient vapor return pressure (in the upper R tank) through a small transfer pump and partially open shutoff valve to the left 1/2" fuel line halfway between the Left tank and the selector valve.

 We believe that hot soaking the fuel on a Gulf Coast afternoon combined with a rapid climb to altitude and this small source of vapor at the lowest pressure part of the line introduced bubbles into the flow and cavitated the boost pump (set on low boost). We doubt that a switch to high boost would have corrected the problem. It did clear after 5 minutes, after the tank was switched and the small air leak removed. There is no connection between the pressurized cabin and this source. It comes from the right vapor return line to the left fuel line.
 We were able to duplicate on the ground only after 5-6 high power hot afternoon runs simulating takeoff at full power. Though ambient pressure was higher at sea level than at FL220, the rumbling down the runway probably shook the "coke can" a bit and allowed us to duplicate the Fuel Flow interruption.
Thanks for all the help from the many responders. I plan to review the Reid Vapor Pressure (RVP=44 KPA for 100LL) implications here and get some data on what's really happening in that very low pressure part of the system.

Detail:
 On Friday, while waiting for parts and the inspected engine driven fuel pump,
 we examined the filters and fuel flow
again. Since the flow was not routed through the FF sensor, we could only
estimate, but flow seemed higher out of the left than the right tank, with intermittent flow and possible air bubbles. Looking
further, I recalled a modification. There was a (unused)
fuel transfer system for an "aux tank" that had been "capped
off". We wondered if the combination of a slightly open valve and a leaky
cap might explain the problems. We found the valve slightly open and there was no cap at all.

We discovered the flow path was not capped, as we
had all been informed, but was set up as a transfer from the left
to the right tank. Fuel from the left  tank 1/2" feed line to the
selector valve also supplies fuel through a T into the transfer pump,
then through a shutoff valve to the right tank's vapor return line!

Question: Anybody see this config before? Why would the earlier builder set it up
this way? It seems backward for the fuel transfer scenario from an aux tank in
the back seat, because of the T into the left fuel line.

In any case, this open path from the right vapor return to the left gravity
feed line to the boost pump solves the mystery of intermittent fuel flow. Why no other occurence in almost 200 hrs?   Usually it was cooler, and the stepped climbs took much longer because of the baffle seal leaks. We had to allowing engine cooling  in stepped climbs, which likely cooled off the fuel too and helped avoid this event. (Remember this all started on a day when we were very happy to have achieved far better CHT cooling on climb!) We had experienced slight roughness usually at night or in weather or over water, but thought it was corrected by the low boost on above 10k procedure