Efficiency cores would burn less energy for any amount of workload, provided the delay is acceptable.

The efficiency cores may be always on, or close to it, so the high performance cores being turned off for longer saves energy too.

Really depends on the architecture in question though. Not very familiar with M1.

IIRC the efficiency cores are usually always on¹ with the more power hungry ones being spun up when needed². I'm being ThatGuy and have not actually read the article, but I assume they are somehow encouraging the scheduler to keep these tasks on the efficiency cores despite CPU being their main bottleneck to avoid them causing power cores to be woken up so energy is being saved this way³. It could also make other tasks perform better by leaving the power cores free for them to chew, essentially similar to nice-ing the process.

¹ Obviously this may vary by architecture ² i.e. if all the low power cores are running close to capacity due to multiple active processes, or any task is spinning one or more efficiency cores near 100% for more than a few jiffies so is likely to benefit from being moved to high power cores ³ Blocking the “this is a CPU bound task, maybe I should engage the other cores” taking effect for this process/thread

Modern cpus do much more work than needed when executing code (branch prediction and speculative execution).