The current wisdom with image handling and workflow appears to be heavily weighted towards preserving all image processing steps. I thought I would come in with a contrary view.
Something I learned when I trained as a carpenter many years ago was - measure twice, cut once. This bit of old wisdom is not only about professionalism, it is also about speed, efficiency and conservation.
My experience with the preserving all steps approach to image processing is that it can too easily degenerate into - I'll try this now and I can always come back and fix it later. That might be useful while you are experimenting and learning but long term the objective has to be to get it right first time. Why do it again?
Sometimes you see an answer to 'why do it again' - 'because my skills will develop as will the technology. I want to be able to come back and redo it better later'.
I don't disagree with the thought, only with the conclusion. If your skills develop, as does the technology, how much of your original work will really be useful next time around. Say Raw Processing improves by leaps and bounds. Or perhaps there are huge improvements in the printing of wide dynamic range images.
Will anything of your original multi-layered image be of use to you? Not much is my guess.
Also, how much of your time are you likely to divert to reprocessing old images when you will have a new wonder camera giving you immeasurably better images in the present. Sure, there will be some you may want to rework but the proportion will be small.
With this in mind I would say develop a workflow that delivers you current best practice now and commit to it. Do not preserve every possible layer of your workflow process. Keep what you really need and the rest - get it right and commit to it.
Let's take a contentious example. Say you do a strong final 'print' sharpening. Many experts advocate creating multiple sharpening layers and applying varying levels of opacity to each layer. Is that really best practice? I can't see that it is likely myself. Think of what you are attempting to merge wth these opacities, one possibly oversharpened set of edges with a second rather blurred set. How likely is that to be better than finding an optimal level of sharpening for your purpose and accepting it at 100% opacity.
There are major archiving considerations also. With my 20D a 16 bit raw image comes in around 10Mb. As soon as I process this to a 1 layer tif I am looking at 46Mb. Because of the way tif's work, a second layer trebles the image size to around 130Mb, even if it is only an adjustment layer. Before you know it you have images taking up 250 Mb on your Hard Drives! Does that make any sense from an archival perspective. Barely at best.
I would advocate the following:
1. Spend a lot of time getting on top of the techniques - so you can produce best practice images now.
2. Use automation to the full - instead of preserving every possible stage of your workflow, know that you can recreate the image from scratch quickly, optimally and efficiently.
3. As far as possible aim make your workflow exact to meet your requirements - avoid fuzzy processes such as partial blending except when this is really contributing something to the quality of your final image.
4. While you are actively processing and printing images by all means use multiple layers where they are useful.
5. Long term, archive the Raw original and perhaps a couple of flattened processed images, or perhaps even the odd layer which involved painstaking work eg dust removel. But don't try and keep everything, it is not going to help you.
That will do for now.
Andrew