There are some scenarios that DISM doesn’t work in as you would expect. When I set up a reference image for a new platform, because of my imaging topology I have images that are pre-sysprep, post-sysprep and post-dism. Generally I would choose a pre-sysprep image, but then it has to be mucked around with and sysprepped before it is ready to use. If I only had one laptop, maybe it would seem simpler to use the post-sysprep image?
Well, it seems not all images are equal. We got an Elitebook 8540 laptop, it’s the only one the school will ever have. So I thought that a post-sysprepped image with the use of DISM to give it the right drivers to start up would work. I did something similar last week with a 6710b and used the pre-sysprep image because it would need to be cloned. The problem is that for my 8540p, the post-sp image doesn’t work. It bluescreens before the Windows 7 animated moving pieces of logo have even finished circling. But if I use a pre-sp image exactly the same way, there is no problem.
So it appears there is a good rationale for having three image generations – apart from the sysprep limit that forces us to have pre-sp images in the first place. Perhaps there is not a need for a separate post-dism generation (the dism stage is done on the post-sp image to prepare it for final deployment). At the moment the process for readapting an image to a new platform is to copy the pre-sp image from another platform, dism it to remove old drivers and add new ones, and move the serviced image to post-dism for deployment prototyping. Once the image has been prototyped on the target platform there is then a series of pre-sp, post-sp and post-dism stages for the reference image for the new platform. In this case some of those steps might be skipped.
So it would appear there is some difference between an image that has been sysprepped and one that has not when it comes to servicing the image offline so that it can be reused on a different platform.