Hi, I know you posted this a year ago and I am hoping this comment finds you well. I first found out about Adapteva and the Paralella probably a little before you posted this article. I stumbled upon it as I was looking for stand-outs in the single-board computer space. My personal needs are for a general purpose ARM board, but I was looking into clustered SBC's simply out of interest and stumbled upon this very unique board.
I'm actually quite surprised that this board hasn't received a lot more attention considering the huge move towards machine learning and high-volume data processing. Sure Nvidia's Jetson platform is awesome, but at quite a premium. If I understand the Paralella's capabilities correctly, the Epiphany co-processor is like a dynamic"cluster" of individual CPU's, that are like a FPGA, but running as a slave of the ARM/FPGA which acts as a controller that can be re-configured to handle the dozens of workloads besides ML, such as soft-defined networking?

I don't work on this stuff directly, but I spend a lot of time reading it simply out of interest. From reading the Epiphany Ref, it seems to me that Linux runs on the ARM A9 as the host OS, and on the Epiphany cluster each nodes is a custom or semi-custom "OS". Almost like images running in a docker swarm.
I was wondering, if it would be possible or even beneficial to use one of these boards with Inferno OS? You may already be familiar, but Inferno is Bell Labs' last OS and evolved from Plan9, still a successor to UNIX. It's a distributed OS using a very light VM at it's core, it's able to run bootstrapped or natively. It's purpose was to make hardware-agnostic and transparent systems from pooled resources. Regardless of Inferno's relevance to Epiphany, it seems to me that there are many projects that would see insane value and applications of this SBC. Would love to hear back from you to know if I'm understanding this correctly or not. Cheers.