Recommendations for advanced linux system administration training?

Red Hat has various training courses for many levels of prior ability, including multiple classes in the categories of http://www.redhat.com/training/paths/core-system-administration.html and http://www.redhat.com/training/paths/advanced-systems-administration.html skills. The main page is http://www.redhat.com/training/. Full disclosure: I work for RH, but not in that part of the company. I'd normally avoid pepsi-blue here, but it seems like a fairly direct answer.

I have delivered Linux Professional Institute (LPI) courses, it is a good syllabus and will take you through the things you ought to understand. Self-taught system administrators are like self-taught drivers. So I long as I don't have to share the road with them I have no problems. The trouble with your piecemeal approach is you may not appreciate the side effects of what you are doing and may introduce problems which lurk for a while then bring servers down in the middle of the night. Anyway, how are you going to Google for it if the network is down? Google can't teach you a disciplined and methodical approach to fault diagnosis (neither can LPI, a problem with all these certifications). The following books should be on your shelf, "Linux Administration Handbook" by Nemeth et al, "The Theory and Practice of System and Network Administration" by Limoncelli et al and "Linux Quick Fix Notebook" by Harrison.

I second the Gentoo suggestion. Consider it a sort of 'guided' linux from scratch - which is a great, somewhat accelerated way to learn what-does-what-and-why because YOU're the one setting it up.

Another good place to firm up your toolset is with the command line toolset. Knowing a little bit about the shell utilities will help a great deal when you're poring through log files looking for the needle. You needn't be an expert, but at least a passing acquaintance with the likes of sed, awk, cut, uniq, wc, sort and some of the others will server you very well. Basic shell-scripting is also a very useful skill to have, and if you want to make it a bit more challenging, stay with the Bourne shell (/bin/sh) rather than BASH. Generally, you want to become something of a toolsmith. Having a box (or VM) of your own to pound away on is probably the best way to keep your chops up, and any old x86 bit of hardware will do. More detailed troubleshooting tools: http://en.wikipedia.org/wiki/Strace (for peering into running processes) and http://en.wikipedia.org/wiki/Tcpdump (for spelunking into network traffic). Between those two, the shell stuff listed above, and a decent central logging facility (you are logging everything to a syslog server, yes?), you've got a fair bit of what you need to troubleshoot almost anything.

What are your company's core competencies? What are your own core competencies? I ask, because the best answer may be to come up with a more formal arrangement with your outside SA resources and your clients about response and resolution times. When there is a problem, you see if you can fix it yourself within X amount of time. If you can't, you bring in the big guns. This doesn't preclude additional training on your part, and, you might even work something out with your outside resources for training, and/or consulting on non-time-critical issues in a way that lets you learn from them.