It doesn't really matter what the configuration was before the formation of the horizon; what matters is what comes out again when the horizon shrinks due to BH

It doesn't really matter what the configuration was before the formation of the horizon; what matters is what comes out again when the horizon shrinks due to BH evaporation. The problem exists classically as well as in QFTs: if I drop a trillion particles of energy e into a black hole or a two trillion particles of energy e/2 with pairs stacked up on top of each other instead, how do I tell the difference? With no-hair, I can't. If the horizon persists into future timelike infinity, I don't really care about this -- the information distinguishing between these two cases is hidden behind the horizon and I can shrug it off. The same is true if there is a remnant rather than a complete final evaporation. But if the horizon goes away altogether, I may need to distinguish between the trillion and two trillion particle cases; that's especially true if I care about unitarity, as I should when using a QFT. The main problem here is the singularity, and the inevitable collision of the trillion or two trillion particles with it. If I can abolish the singularity so that these particles evolve normally within the horizon rather than being squashed into a singularity, my information loss problem goes away, and I don't care what happens to the horizon at all. A useful high energy theory of gravity should solve this, whether it does it by changing spacetime, or fundamental particles, or produces a working renormalization (one could conceivably defang the relevant infinities in principle, analogously to how asymptotic safety, but not necessarily using ASG's nonperturbative renormalization), or some combination, or something else entirely. The retreating horizon is another problem, but any approach there has to deal with at least stellar (and potentially smaller) mass black holes and also supermassive black holes. The former may evaporate entirely in finite time, and one has to be careful about tricks that prevent that somehow because the latter may have arbitrarily flat spacetime at the horizon. These features tend to break remnant and RG approaches.