If Systemtap is so great, where is a similar document showing how Systemtap has been used to solve problems in both production and non-production systems?

At this point, the people saying that Systemtap is equal or even better would be the equivalent getting advice on which beef is the best from vegans and the sales people trying to sell you a cow.

In short "Where's the Beef!"

At this point I could care less about who has a better license, etc. I want more ZFS style options out there, having inline data redundancy/checksumming and corruption tolerance with the ability to scale and be device agnostic is fantastic. It would almost be enough by itself for me to switch to OS X as my desktop OS. Worrying about data corruption (which has happened to me a lot lately it seems) is something that I would like to do without.

We could probably do a whole 'nother thread on ZFS and it's pros and cons. I have heard nothing but high praise for ZFS from those who are familiar with it, though I have also heard from some Linux people that it violates some architectural principle for them in terms of layering of abstractions.

Having been in the real world a few years, I can say for sure that people's intentions often don' reflect reality. The reality is that there are plenty of applications which started life as a prototype, but which went on to become the production system without being rewritten.

Most often, these systems go into production and later on as more load comes in, there are strange performance problems or bugs, and there is no good way to instrument live code on a production system to see what's up.

This is where DTrace, built into Solaris, really saves some people's bacon.