I'm in favor of trying many different ideas, for a while, i.e., when a problem is new. But at a certain point you need to address issues that are viewed as critical. In the academic fields that I'm familiar with, I don't see that occur as frequently as people try new ideas that don't obviously address the most critical issues. These new ideas typically do address some issues but they might be regressions in some areas, ignore other important issues, or address the issues they focus on poorly.

In software development you have bug trackers. In scientific research you have review articles. These are useful tools that should be used to guide new work. Otherwise you're not doing simulated annealing or whatnot, you're just doing an inefficient random search.