I work with documents mixing text and bitmap images, for which the final format is a PDF file. Most of the times, I produce my own graphics, which means that I have a choice of:
Creating all bitmap images at very high resolution, embed it in the document, and post-process the final PDF document to downsample all images to my target resolution (say, 300 dpi).
When I create the bitmap images, export them at a size consistent with the target resolution and intended size, then include them in the document (and no need for downsampling later on).
The second option has one major downside: it means that if I change my mind later (e.g., I want to include an image at twice the original size), I have to recreate the bitmap image file. It's not a very big deal, but still. Because, I cannot see any downside to the first option, that's what I regularly use.
My question is: how do you arbitrate between these two options? Do you always work with overly high resolution images, and downsample everything as the last step?