So far it was all just words, and I wanted to demonstrate this, so I ran a test. Test consisted of a 50x50mm cube printed 5mm tall. I printed it with two perimeters. I duplicated it, and set up different processes so I could test two of these cubes side by side during the same print.
All layers were .2mm high.
Test 1:
one cube printed with 10% infill, 4 top layers
one cube printed with 10% infill, 3 layers of 30% infill, 4 top layers
Test 2:
Same as above but 5 top layers for both cubes
Test 3:
Same as above but 6 top layers for both cubes
Looking closely at the test cubes, I'd estimate that the ones where I added three 30% infill layers before the top were about two full top layers ahead of the cubes where I just printed top layers over 10% infill as far as establishing a good, solid top layer with little to no trace of underlying infill. Had I gone with a 40% infill for three layers instead of 30%, it would have been more dramatic a difference. I may run another test and just print the same cube with three layers of 40% over the 10% main infill and four top layers just to see how it comes out. Due to lighting and using a cell phone to take these photos, the effect can be somewhat exaggerated in the photo, but even at four top layers, the cube with the 30% infill before the top has the underlying infill texture very visibly yet in terms of the actual surface texture, it's very faint differences in the surface, whereas the cube where the four tops printed over 10% infill the centers of those cubes where the infill is are far more undulating and bigger actual surface differences.
This was all done in PLA. If I'd used PETG the differences would have been far more dramatic, since bridging of the top layers over the infill would have taken an extra two or three layers to firm up.