That is indeed a very interesting discussion, and I know for a fact that the majority of our users actually uses the advanced output as you do. That doesn't necessarily make it the best way to do it though.
I would say that if you have a need for having the advanced output open (or for midi/osc/dmx control over the advanced output, same thing really), you are using the advanced output to position or control the opacity of content. Which are per definition *compositing* actions.
I think that unless your stage is physically changing shape, there shouldn't be a need to change anything in the advanced output. The advanced output is used once to adjust your output for a LED processor's pixelmap or to align content to a physical object.
Positioning of content, opacity of content, anything to do with content really, are *compositing* actions and should happen in the composition.
If you don't do this in the comp but via layer to slice routing, you end up in all sorts of dead ends, as you are well aware. These dead ends in turn lead to all sorts of feature requests like effects for slices, blend modes for slices etc. These feature requests are basically the result of an XY problem
Let's define X as follows: the user needs an easy way to position content and to easily make changes in this positioning.
The first half of X is easily done in the AO, but the second half isn't. This leads to the request for what the user thinks is the best solution, namely Y: external control over the AO (and by extension effects and blend modes in the AO). This is indeed a logical request once you've headed down the dead end of positioning in the AO. It's also how for instance Madmapper solves things, which makes sense because they have no friggin' compositing stage. Of course they would add blends in the AO, where else are they going to put them? So I can see how this request can come up, it's just logical thinking.
But, but, buuuuut....
If you take a step back, a far better solution would be to actually give the user X, which is better tools to position and control content in the composition. Which is the thing that they are actually trying to do.
If X can be done easily and intuitively, there is no need for solution Y. The user gets all the benefits of working in the composition (blend modes, a preview that makes sense, a layer stack that makes sense, global effects that are actually global, etc etc). They also get all the benefits of easy positioning and changing setups, because these can be treated as clips or layer effects. They can be turned on and off, like all other clips and effects, without the need of having a second window open or to remember a midi map of something that you can't even see during live performance.