What's the probability that sometime stealing a stereo out of a car is armed? Five percent? Twenty percent? Fifty percent? Go ahead and try to quantify it in your mind.
Because it matters. Here, Osborne's working on his own car, but the police -- who don't initially know whose car it is -- wrongfully think he might be stealing the stereo. So, after they approach him, they pat him down.
Assume the initial detention (telling him to "Stop" and "Come here") was justified. With respect to the patdown, the law is that the police can pat you down if they have a "a reasonable suspicion that you are presently armed and dangerous." Does the fact that you might be stripping a car or stealing its stereo make it reasonably likely that you're armed?
Justice Sepulveda hints in this opinion that the answer is "Yes". Holding that even though "[a]utomobile burglary is not  recognized as a classic violent felony . . . . an individual suspected of such a crime [may] reasonably be anticipated to be armed with a weapon (such as a knife or a firearm)." Justice Sepulveda also cites with approval secondary authorities in support of "so-called 'automatic' patsearches where an individual is suspected of burglary." The theory being that if you're stripping a car, you're likely to be armed.
Ultimately Justice Sepulveda adds a couple of other (incredibly minor) factors here that arguably assist in establishing the fact that the defendant might be armed -- including, ironically enough, the fact that Osborne was a large 240 pounds (though to me that proves that he was less likely to need and carry a weapon) -- to hold that the "totality" of the circumstances suggested that he was armed. But clearly the biggest factor was that he was allegedly stealing a stereo, and I have a keen sense that that alone would have been enough for the panel. Which is an interesting result. Think about whether you agree.
I have a separate thought about post-hoc conclusions in these search and seizure cases that I'll briefly share as well. Almost all of the cases that get to the Court of Appeal are -- not surprisingly -- cases in which the search does indeed result in evidence of a crime. For example, here, there indeed was a gun. Searches that result in evidence often reach the Court of Appeal in the context of suppression motions; by contrast, searches that do not result in evidence rarely go up (since there's often no arrest, conviction, or suppression motions), with the rare exception of Section 1983 cases.
Not only does this lead to some results-oriented jurisprudence -- e.g., you want to affirm because we know the guy's a bad dude, plus we naturally give deference to the officer's suspicion because it turns out s/he was right -- but I think this one-way rachet may also tend to give appellate judges a somewhat distorted view of the facts and circumstances on the street. When you've got lots of cases involving searches that turn up guns, you may have a keener sense that virtually everyone has a gun (or that most people searched have a gun) than actually exists.
I don't know how much this plays into things, but I'd be surprised if there wasn't some tendency along these lines. It's only natural.
Of course, someone could rightly respond that someone sitting on a bench -- or, like me, in an ivory tower -- may similarly not recognize how prevalent guns really are, since none of our friends are typically packing on the job. Fair enough. It'd be interesting to figure out how these competing factors actually play out, and which one predominates.