I guess there is probably a large community of people who only share publicly projects they consider portfolio Worthy. I have found sharing valuable in both failed and successful projects when interviewing, teaching, and spec'ing a new project.

You're correct in noting that some of these projects fail, but as long as there is sufficient code the projects they can often be interesting.

Sort of side projects that I pull together in another languages are often to explore language features that are not readily available in my common tongue. Certain languages have made completing certain tasks far easier like heavy Matrix programs in Octave or Julia, n-body simulators with extended unit types in Julia, multi-threaded cryptographic toy project in erlang, statistics problems and visualization in R, easy concurrent types and strong explicit typing for model train controller and Ada. I have also used a new language for a interview coding challenge (for fun).

For example, I've ported my Proximity game a bunch of times. Some of these never got completely finished and released:

* Actionscript (Flash, released and playable online)

* Java (J2ME and later Kindle -- the Kindle version looked awesome, too bad Active Content is no longer a thing)

* C++ (Popcap framework)

* C# (XNA, released on Xbox 360 on Xbox Live Indie Games)

* Objective-C (OpenGL ES, released on App Store)

* Python (I just used text graphics)

* Lua (Pico-8, almost done but not quite)

* Physical board game prototype, hoping to find an interested publisher

* C# again (Unity this time, and making it in 3D)