Right - what I'm imagining is, you SIGHUP the supervisor, and that causes it to tell the children to die, waitpid() on the child processes, then it re-exec()s itself (the supervisor).
The new supervisor instance, then, goes about its normal business.
exec()ing yourself means that Python would e.g. reload all the modules.
for real
sounds like a good solution
definitely interested in that
create an issue on github