Thank you, I will look into it. It could be the OOM killer stopping the Julius process; I'm not sure pyjulius knows how to handle that gracefully.

Everybody: It appears that I made a mistake in Git when I merged this commit - I overwrote my changes to the navigation code. As this only affected the upstream files, I did not discover the issue on my testing devices (which still had the correct code); however, my build server pulls changes from Github, and built the package with the wrong data. As such, turn-by-turn navigation is completely broken in the package in the repos (it will just say "<destination> is <distance> away" and then hang forever). I will upload a patch this evening, once I can verify that that commit did not break anything else.