I DO NOT recommend building Linux from Scratch, or going say Gentoo way, as a way to learn Linux. It is a way to learn how to build a Linux system.
I think that it is imperative that cranky newbie Linux users do indeed try LFS. Only then will they appreciate the work that goes in to preparing a downloadable ISO.
The double-edged sword is that there are many who think that LFS is as easy as a pre-prepared distro (until they try to do it). And when it doesn't work out for them, the whole internet needs to hear about it.
I disagree, because after all, it is just a tool, and you don't need to be aware of the effort or cost that went into the tool to wield it properly.
(Which is basically what Mechatrommer said above.)
The possible workflows differ a lot. Some users do not need to know the OS details at all. They might even have good grasp of Bash scripting to batch-ify repeated workflows; and that is right now quite portable between Linux, BSDs, and Mac OS; possibly even partially Windows (although I don't know if/how Bash and native Windows apps mesh). With just a few command-line tools, say
inotifywait on Linux,
fswatch on Mac, you can do magical folders (that automate format conversions and other such stuff, to another folder), saving a lot of time in certain work situations.
Some users can benefit from OS customization. For me, small things like personalized splash screens (my own one is Tux rolling his eyes in opposite directions) make it
more mine. An organization certainly benefits from their machines having some such visual details (it changes how cow-orkers treat their machines; they're more willing to consider, ask, and invent better workflows), and the sorts of customizations a capable Linux admin can do. However, those admins rarely have done any of the work the people there do, and therefore cannot usually imagine what kind of changes would make workflows better. So, if you have users who understand the basics enough to have suggestions, or at least identify slow/problematic parts of their workflows, they can describe them to the admins in a way the admins can actually act upon.
(Whenever I've talked about Linux to organizations, I've always first done some informal interviews, finding out exactly
how people do their jobs. Which kind of added to my depression; it is amazing how much unnecessary stuff and extra efforts people spend in their day-to-day work. Easily makes one lose all hope for humanity, if you do it too much, really.)
In other words, I don't even think that ordinary Linux users
need to be able to customize their own workflows; but I do believe they need rough understanding enough to describe the problems they are having, in a way that those with the skill (but not the experience with that particular workflow) can suggest and implement the changes.
See? There are lots of different "depths" of knowledge, and even a rough understanding is extremely useful. Not everyone needs to be a Linux admin, but everyone should have enough understanding to be able to describe their problems in a manner that can be acted upon.