I write software that scours the internet for porn and other NSL content, then sends it to school kids.

A Raspberry Pi 3 Model B

It's connected to my 3d printer and runs octoprint allowing me to upload print jobs. and control the printer from my home network.
It serves up the Pi camera video stream.
It can also switch the printers light on and off.

No cluster setup.

This seems strangely just a little bit wrong.

As mentioned there's no units on STL files so the slicer assumes millimeters but warns you if it thinks the scaling is obviously not right.

My blender units are meters.

If I just export the default cube from blender (which is 2x2x2) and load it up in the slicer I get a prompt saying the units seem to defined in inches. It's assuming I can't possibly want to print a 2mm wide cube. Do I want to convert it to inches?

If I answer No, I get a perfect 2mm cube, good luck printing that. If I choose Yes, to convert it then I get a cube 50.8mm across. (2x 25.4mm) exactly 2 inches as expected.

Going back to blender, deleteing the default cube and creating another cube but scaled to 0.1m. This time exporting to STL on the export save dialog there is a scale setting: set this to 1000 (to convert from blender 1m units to Prusa 0.001m units). Now the STL is imported into the slicer with no fuss and is exactly a 100mm cube.

It seems I can't recreate your problem.

If you had this much buffer memory what are the reasons to have swap space as well?

Many programs do stuff once during startup that they never do again, sometimes creating redundant data objects that will never get accessed in the configuration its being run in. Eventually the kernel memory manager figures out that some pages are never used but it can't just delete them. If swap is enabled it can swap them to disk instead. It frees up that RAM for something more important. It's usually minor but every few MB helps.