Could it be designed to limit variance and the impact of clever hardware and its unpredictable nature? I can imagine benchmarking on a new high-end desktop machine would require a lot of work to make sure your benchmark isn't impacted because you ran it Friday at 11:44 and the cron job that runs at 11:45 messed up a few key caches, made your CPU heat up that extra 1 degree celcius causing some throttling to kick in, causing your prng to be seeded with a value that triggers a few worst-case performance scenarios etc.

There are a lot of factors that can influence a modern system and most of them seem hard to control. For a 8086 you could probably account for all these factors if you put in the hours and run it in a highly controlled environment but I doubt its even possible for a modern CPU, let alone the whole system.

Of course this isn't a problem for most benchmarks as most of them are only useful for demonstrating real world use-cases where the variance is expected and the goal is not to benchmark the performance of a single element of the system but the perceived performance you can expect in a real-world scenario.

Or its simply designed by someone who loves this kind of system and found a good excuse to put one to use.