@witty-crayon-22786: I had another thought about it. The potential performance regression caused by the patch is the same one we’d get if we used cache_target_dirs as it’s currently implemented with compile