Agreed. "Magic" is just another word for "I don't understand how they did this". Whenever you meet "magic", you should dig into the code until you do understand "how they did this".
In particular, there is a comment[0] that I think is particularly worth reading:
There are two different definitions of magic in play here.
1. The use of advanced language features in a manner that might be confusing to a novice programmer.
2. The use of ANY language feature to automate some process that a programmer would not expect to be automated.
[0]: http://www.b-list.org/weblog/2009/jul/23/magic/#c118751
The real question for any interesting application is not the how or the what, but the why. You can't answer the why just by reading the code. You need to understand the problem and the trade-offs, and for that you need comments and documentation.
I think some people feel like rails went a little overboard on things like this.
A few of the "magic" things I've encountered thus far in Rails can be easily demystified by looking at the code. However, like magic shows, knowing how the trick is done doesn't mean that I can't enjoy the spectacle of it. The execution of the trick is every bit as important as the mystery.
I'd put forth this definition, which applies to both magical performances and magical code:
Magic is taking something which is traditionally considered impossible and making it look effortless.
Some magic tricks really are effortless. Others require painstaking setup and flawless execution. Same thing with code, but only the former is really valuable in production systems.