I must add that the downside of the method is that the preprocessing has to be done in "finite" bit depth, basically clipping the peaks again, but in a more gentle way, if not normalizing the material so that the peaks are untouched (and could be in very high level compared to the rest of the song). If ST makes the declipping, it doesn't matter how high the peaks reach since there's plenty of headroom internally.
So, I must correct my own comment by saying that even better way would be to tag every file / show with some data that points if the file/program is in need of declipping, and then remotely disabling/enabling the filter according to the data. But that workflow could be harder to make than preprocessing