This. I think what you do in that break really matters. If it's taking a sip of coffee, browsing HN or answering a question then it's not going to help.

I have learnt to race cars over the past 5 years. I can lap and lap and lap, but the greatest improvement comes from the breaks between sessions, which can be minutes, hours, days or weeks. I stumbled upon the most effective strategy the other week when training in my simulator: I was doing a small number of laps, say between 7 and 10 then returning to the pits. I'd reflect on how the "stint" felt for a couple of minutes and then have a look at the telemetry, comparing laps and correlating my memory of the on-track experience with the sensor outputs that I could see in the telemetry. After doing this for 20 minutes, decide on some small setup tweaks and jump back in for another 7-10 lap stint, rinse and repeat.

I spent maybe 3 hours doing it and drove maybe 50 laps during that time (I could have done over twice that number if I hadn't taken my approach), but the improvement I got from it by the end was far and above anything I've tried before.

I can also relate it to another time in a real car where I was learning a new track and car in a short 20 minute session. I had done ~15 minutes on track and the session was stopped. I got back to the pits and sat in the car. I asked my engineer about the time I'd done in relation to other drivers and just sat there in the car thinking through my laps and where I could be "better". I went back out on track for the final 5 minutes and made a huge improvement, finishing top. It wouldn't have happened without that brief period of reflection.

TL;DR Do what you're doing for the shortest meaningful period of time. Stop and reflect on what you've been doing and what you could do differently/try/better. Let your mind relax whilst you're doing this - it shouldn't be an intense process. The go back in and put your reflection into action.

It's important that you try to make adjustments to your performance, even if they fail. If you just do the same thing over and over again, you're missing the opportunity to diverge from your current behaviour, which is probably suboptimal given your experience.

What you're describing here is really what Ericsson refers to when he talks about 'deliberate practice' (from the 10 000 hours meme). You correctly recognized that instead of doing more laps some of your time was better spent on deep analysis.

If we stick with the racing analogy I think what the article talks about is that you would've seen solid gains if you had tried to run a single lap perfectly, stopped, stayed in your car, and just relaxed while your subconscious got some time to process everything that just happened, without your bombarding it with new information. If you were to follow the instructions to the letter you'd hang out for a time equivalent to your lap time, but they left it as an open question if this mechanic scales beyond the 10s repetitions used in the study.

I hope they do more of the studies because the required time of rest could plausibly scale in different ways:

1. Your subconscious can only remember so much stuff to replay and learn from while resting so a break beyond X isn't beneficial.

2. The rest time needed is linear with the stimulus.

3. Long action sequences are such a rich source that the subconscious can mine them for a long time. Beneficial rest time scales exponentially.