True story: Some years ago, a consulting assignment involved analyzing a bunch of data saved on a slow as molasses NAS using SAS's data format. The work was CPU intensive but IO throughput was the bottleneck. The analysis was close to being embarrassingly parallelizable[1] given an appropriate partitioning of the data.
The only thing installed on the remote system I was given access to via RDP was SAS whose macro facility tends to be cumbersome. Since the original CSV files were also stored on the same NAS, I asked if could have any programming language installed in my home directory (there was absolutely no way for me to upload anything, the system had no internet access other than to allow RDP, and I wasn't going to violate any rules.
The sysadmin head read BOFH for career development in the 90s, so, the answer was a solid no. I had to only use what was available on the system. The problem was simple: Each run took about two days: Start the thing at 7 am one day so you can hope to have the results before 4 pm two days later kind of thing. That's when I remembered `cscript`.
Using that, built a "thing" that orchestrated the copying of data in 100 GB chunks to `%TEMP%` which was still on the precious 256 GB SSD :-), launching multiple jobs etc. Task manager screenshots: https://www.nu42.com/2013/12/happiness-is-making-most-out-of...
In less than a week, jobs were running in four hours. sysadmin remained happily oblivious. Had to explain what I had done to the rest of the group.
[1]: https://en.wikipedia.org/wiki/Embarrassingly_parallel