What in the heck is going on with Garage? I'm most interested in the non-parallel 1GB download speed.

Garage 75 MB/s vs rest of the middle of the pack of 200(ish) MB/s. Is this just language choice? There has to be some poor tuning of code, settings to adjust, and libraries in there.

Would be nice if the article listed the specs of their test machine: cpu, memory, disks, and network.

There's no source code, they copied files back and forth in a completely scientific and repeatable way.

where is the source code for this benchmark test?