Google Reader had become amazing at getting updates virtually as soon as they were published. I've been comparing how well the competitors do, and some of them are more than 30 minutes later than Reader at finding and showing stories.
I also imagine it's because google was requesting the feeds more often than others, or leveraging the feedburner system to know when they were updated and other such things.
It handles all the "dirty" parts of feed reading (checking for new items, normalizing fields) and adds near instant updates for sites that ping hubs.
When I finally get around to writing my own feed reader, I definitely plan on using it.
I wonder how hard this would be to get it implemented in some of the other readers.
http://googlereader.blogspot.com/2005/12/xml-errors-in-feeds...
That is to say, I completely agree with you. Google Reader had an excellent backend (being used by 90%+ of people meant there was a virtuous cycle happening), and I especially noticed this when I developed a prototype backend that I never actually used. Deciding how frequently to check feeds is complicated.
In the end, given that this is a side project, I'm never going to be able to compete with the big players, especially at scale, and to be frank, I don't find this side of it as interesting or enjoyable to develop.
The good news is that this frontend could be modified to work with another backend if there was a good reason to do so.