TL;DR: auto-generated denser infill layers prior to printing top layers over thin air would decrease the number of top layers that must be specified in order to successfully bridge the infill air gaps and get completely filled, non-drooping top layers, saving both printing time and filament usage.
I actually came to the forum to post this idea myself. I saw your post, so I'll wholeheartedly support your request instead.
I already do this manually using processes, which is a bit of a pain. In order to reduce the number of top layers that I have to specify in order to get a good, solid, droop-free top, I'll create a new process at the level where I want to introduce a denser infill that will better support the top layers. In order to have the denser infill always include the existing "sparse" infill lines, I make my denser infill an integer multiple of whatever percentage I used for the sparse infill.
Just last night I did a print where I was using 8% infill for most of the print. Two layers before getting to the top I changed it to 32%. This made infill lines that fell on the same lines as the 8% infill, with three more lines in each direction in between those. Yeah, the first layer of extra lines sagged quite a bit and many of the lines were a little broken, but the second 32% infill layer improved this greatly, and by the time I got to my top layers, what was there from the 32% infill (which wasn't perfect by any means) was enough to give the first "top" layer much more to hold onto, with the result that my top layers formed up nicely within two layers or so, not the three or four that would have been required if the first top layer were printing over mostly thin air.
Your suggestion would simply automate this. Here's what I envision:
1) User-specified number of "denser infill" layers.
2) User-specified multiple of the sparse infill percentage, ie: user can specify how dense to make the denser infill layers.
It would be a bonus if we could specify a number of density-boosting "steps," as well as specify how many layers to print at each density step. I'll illustrate by example. Let's say they implement this idea, and the rule for going from sparse to dense infill percentage is to simply double the percentage. Say I set my infill to 10%, and specify two density steps and two layers per step. My print is progressing, and I'm seeing a good 10% infill during the print. When the print gets to four layers before a "top" layer, in that top layer region it changes to 20% infill for two layers, then 40% infill for two layers, then prints the top layers. What's the benefit here? Well, in those four layers under the current sparse infill-only scheme we would see 4x10% = 40% of a single layer's worth of plastic printed in the infill. Adding this new scheme would increase the amount of plastic used so: 2x10% + 2x30% (the difference between what it would have printed at the 10% setting and what it would print under the new setting) for a total of 80% of a single layer's worth of plastic. This 80% worth of a single layer would undoubtedly result in a much better foundation for the "top" layers to print on than simply adding one additional top layer printing over 90% void space. You'd probably have to add at last two or three additional "top" layers to equal the effect in terms of establishing a certain number of successfully printed top layers.
One huge benefit of providing this auto-generated denser infill would be that it only densifies the infill in areas that are directly under upcoming top layers. With my process kludge it's densifying the infill everywhere based simply on the start and stop heights of the process. Look at it this way: Say you're printing a model of a staircase. This illustrates the problem nicely. As you approach each "step" in the staircase you'd like the infill to densify so that the top layers for that step form up more easily with the fewest number of top layers required to do so. You don't want the infill percentage changed in the regions of the models that are under the other steps of the staircase that we haven't reached yet. Doing this manually, with processes, would resulting in densification of the infill across the whole model each time we approached the next step in the staircase. This is wasteful of both time and plastic, not to mention the user's time due to having to add and configure processes at each step in the staircase. In short, adding this feature would be extremely useful, generating the best printing results for the least printing time, filament usage, and time spent dorking around with processes and whatnot in the slicer.