Having been stung once or twice, at least, by wrong tank v. engine alignment, I chose to do the stupid, simple(?) thing about it...
My test-bench/break-in stand has a rotating baseplate, parallel to the shaft centerline. A (too complicated...) tank bracket allows adjusting the relationship during bench running.
The Tatone, or is it J'Tech these days? mount on the baseplate - with the tank carrier, can be rotated while the engine is running. The tank can be mounted to simulate the in-flight relationship to the engine in upright, inverted, head-out profile and head-in profile arrangements.
You may have read my "wig-wag" method of checking a model for similar or identical "run" upright and inverted? I can simulate those gyrations on this bench.
It works for me, to at least 90% effectiveness. As I often mention re: the "wig-wag:" FINAL tank position must be proven in flight, but the bench "pre-test" gives me confidence I won't flood or flame out in an early flight.
Again, this is stupid, simple, do-it-until-it's-done technology. I have found that similar engines may need very different "tank height" solutions... And it's not predictable by spraybar height...
Applying the "wig-wag" technique to the test bench: -
The tank carrier is lined up with the engine at the presumed relationship. Engine is started and set to a solid 'medium' run that will respond to tank height. (Tank too high: gravity feed adds to the suction feed, and vice versa.) Position the engine with the tank vertex vertically down. Rotate the baseplate to an angle that simulates upright flight - I use about 45° from the vertex down position...
Note RPM, or engine sound if you don't have a tach.
45° Rotate it 90° to 45° off vertical to the other side - which in this example would simulate loads on the fuel in inverted flight. Note RPM - or listen to the engine. The angle where it runs richer is where the tank is "too high," which will be confirmed by it going leaner in the other angled position...
Stop the engine, shift the tank according to what you observed. Crank it up again, and check the two angled positions. Repeat as necessary to reach a layout where there is no - or at least minimal - RPM change at the two angled positions.