That way you might not become "the greatest X86 assembly programmer" or whatever niche you like, but you might become a very proficient system integrator, which is useful in its own right.
It's worked out kind of weird, pushing me (after a long career) into being a founder, and building a tool for generalizing common problems in systems integration. :) My desire to wear lots of hats finally works! But really? I'm more or less a specialist at being a generalist. And I've learned not to chase shiny new tools, because they're very distracting.
That said, hobbies don't need to be productive and half assing something is often good enough. Yes, a simple wedge doorstop may damage the door, but well sometimes that's just not an issue.
These are two different functional modes. The mastery of acting without conscious thought is very hard to attain. See also flow (Cziczentmihaly) and related topics.
Learning incrementally can help, and I'd structured my own career such that I was doing that. I ultimately opted out of it when that stopped being an option. I was simply spending all my time catching up, and not finding myself (nor, to a very large extent, any of my peers) actually proficient with the New Hawtness.
But that doesn't mean you have to live on the bleeding edge, either! A tech that is proven out, used widely and stable, that's worth investing time to learn. In Crossing the Chasm terms (great book on marketing!), you need to learn early majority tech, not early adopter tech. You can be an early adopter for kicks, of course, but don't pretend it's to be more valuable.
The point is, I haven't had to deal with C for a long time. It's odd to be doing it again.