This is very much on-topic as far as figuring out what to do with valid data that OSM no longer wants. I don’t dispute what you wrote, but it needs some context.
OHM as a whole is dedicated to the public domain unless otherwise noted. OSM is not. This discrepancy is rooted in the project’s differing purposes and audiences.
Historically, OHM has promoted a notion of per-element licensing, scoping the “creative work” to an individual element. For the most part, people avail themselves of this option to add Creative Commons–licensed content to the project. However, some mappers have brought in small amounts of ODbL content from OSM, often of their own original authorship.
From the time I joined the project, I’ve been trying to steer the project away from this approach, because I don’t think mixing licenses in a database is sustainable in the long run. I’m not confident that mappers who copy insubstantially from OSM will stop there before it becomes a problem. And what’s the point of saying the database is in the public domain if a data consumer must filter out the Sphinx or London Bridge or part of the Thames because it isn’t in the public domain?
So personally, I agree with you that remapping is the best approach. In my experience, remapping a feature is often easier than copying it from OSM. In OHM, you can still get away with mapping a long highway without splitting it at every turn lane or mapping a shop without fussing over opening hours (which might’ve changed many times anyhow). If need be, you can sketch in the geometry crudely, just as OSM’s pioneers did back in 2008, with the faith that someone else will eventually refine it.
Most of the time, the time sink is researching the dates, not drawing the feature. And yet even the dates can be sketchy, sometimes unavoidably. This is not to say that the abandoned railroads in OSM are worthless to OHM! Scholarship and discovery has enduring value even if it needs to be restated from time to time.