First, there's value in decreasing the viral load of ALL patients regardless of risk, as viral load is likely proportionate to risk of onward transmission. If it decreases onward transmission, then the math would likely favor its use in all cases, since exponential spread gets costly fast.
Second, we continue to focus on hospitalizations and deaths with increasingly poor data and with no focus on disability. Treatment may decrease risk of long covid in non-"high risk" patients, which itself may be beneficial.
Third, while we should expect paxlovid-resistant variants to pop up, the fact that we prescribe it about 4x the rate of molnupiravir and don't see nearly that level mutation should be somewhat comforting. It seems equally plausible that a new "long branch" lineage would pop up with resistance irrespective of evolutionary pressure from paxlovid prescriptions. If that's the case, it'd be a shame to ration its use to only have it go bad anyway (this is basically what we saw happen with all of our monoclonal antibodies).
Fourth, this entire conversation should be less focused on who we deem worthy of one of our last remaining treatments, and more focused on how we're allowing a very quickly mutating virus to spread virtually unchecked and with basically no investment in any further treatment or prevention. None of us need to be prisoners here if we were continuing to invest in new treatments, better vaccines, improved air filtration and ventilation, and paid sick leave.