I've thought about this often and the only solution I can come up with is to disallow myself to think quickly when solving problems, even when I believe that I should reliably be able to do so[0].

I've noticed that when I take the time to think slowly through problems, although the initial activation energy is higher, that I end up saving time in the long run. Also, I tend to end up with a higher percentage of correct answers. This trend tends to increase with the complexity of the problem.

Yes, you probably lose some time on the easy problems, but I look at it as a method of gaining net time over days, weeks, etc.

This, unfortunately, is tough to stick with. It's tempting to skip the easy steps along the way (and let our System 1 do all the work, leaving System 2 to collect dust).

[0] - This is still a working hypothesis that I will undoubtedly go back and forth on for the years to come.