Often I use ACR to open a a large raw file (say, from D700 or D800) which will be used only for a small web image. Is it still worthwhile to do capture sharpening on the file, before moving it to Photoshop for additional tweaks, down-sizing, and sharpening at final size?
Depending on the image content and the down-sampling percentage, it's more likely to hurt than be of any visible influence. With a proper down-sampling method, the risk can be reduced, so then it wouldn't matter much (if anything).
I'm curious about this, because there is a theoretical argument for low-pass filtering (Gaussian blurring) an image before a down-sampled size reduction. This reduces the possibility of spatial aliasing artifacts from the resampling. So, why should I sharpen the image in ACR, if I'm going to blur it in PS?
Capture sharpening won't have much if any beneficial effect on the downsampled result. Targeted sharpening might help a bit if it improves the MTF curve where there is still detail at the final image size, but the best amount and radius would be much easier to judge on the final output size.
I usually make a PS luminosity blend if sharpening layer at 100% which I can switch off or delete for downsampling, and then add a new luminosity blend if sharpening layer at the final output size. Such a layer allows me also to choose whether I want to change the sharpening regime for enlarged output. Unfortunately the Photoshop algorithms for downsampling are not very good (LR is reasonably good), so I do that with ImageMagick anyway if I want the best result.