I'm always expressing the opinion on tsumego threads that people should do more easier problems and less beating their heads against the wall on difficult problems that they would never have the time to solve in a game. (Actually I think both kinds of problems are valuable but I think people usually need nudging in the easy-problem direction.)
Recently I've been putting my money where my mouth is and going through the "Kyu level testing" problems on GoChild. These start out with totally trivial problems that take literally one second to solve, and even at the section I'm up to now (3.3, in the 2500s) I can solve many of the problems pretty much instantly (in a few seconds).
But as the level of difficulty gradually rises, I'm encountering (a) problems in which I really have to stop and count liberties or read out capturing races, causing me to spend 10+ seconds, or (b) problems in which my two-second intuition is incorrect because of some subtlety I should have noticed and I actually get the problem wrong. And I can tell that these are all situations that still fall into the "obvious at a glance, play the right move in two seconds" category for a dan player.
As I work through the thousands of easy problems in this set, I feel like I'm getting these patterns more into my fingers, so that in an actual game they're both more obvious to me at a glance, and can also be used as "leaf nodes" in my look-ahead tree so that when I'm doing actual real-game calculation I can stop there because the result is clear.
A lot of these problems would probably be considered too simple to be worth doing by many 10k players - but if a 3k player still thinks he has a lot of room for improvement in execution of this level of problem, maybe it will inspire some 10ks to spend more time on these close-combat fundamentals as well.
