Just add some more info. I can reduce the number of startup failures by increasing the container-wait-time, though this is very long.
container-wait-time: 1200000
This is related to the uptime monitoring thread: https://github.com/openanalytics/shinyproxy/issues/517
As well as the restarter thread:
https://github.com/openanalytics/shinyproxy/issues/510
...