You can partially mitigate the risks of single threaded Node by using clustering (although it's still tagged an experimental feature) [1].

Also, these worker threads can be made to respawn on a shutdown caused by - let's say - an unhandled exception as [2].

[1] http://nodejs.org/api/cluster.html

[2] https://nadeesha.silvrback.com/setting-up-multiple-worker-th...