Consider it anecdotal, but I've personally never been burned on any jQuery minor point releases. Going from 1.3 to 1.4 and 1.5 has always been cause to revisit my code, but generally, minor point releases have always improved things for me.
You lose almost all of the caching benefit when you use the "latest" version references, like 1.5. They're (necessarily) served with a very short expires header, as compared to the specific versions' +1 year.
That's a very valid point that I hadn't considered.
I'd be curious to see how many people use the latest vs. the specific versions. Excluding the shorter cache on latest, in theory at least, you get the greatest cache benefit from whichever is the most linked-to version, yes?
E.g., if 1.5.1 is 300% more popular than 1.5.2 at a given point in time, then you're far more likely to be primed for any given user, yes?
There's danger in trying to out-think other sites, holding back on upgrades with the expectation that everyone else will do the same. If everyone did that, no one would ever upgrade at all.
On the contrary, I think that the sites which matter (those with high traffic and broad reach, which are also the most actively maintained) will probably upgrade fairly quickly.
How short is the header? While 1 year+ is nice, if 1.5.x is served with, say, a 24 hour expiration header, you can still get a lot of cache benefit if the user visits other sites that use jQuery that day.
