OTP application callback for Snippy.
Starts a :one_for_all supervisor with the ETS table owner and the
discovery store as siblings.
Restart intensity
Web servers often run hot enough that a brief burst of transient errors (a flaky NFS mount, a momentary DNS hiccup during a reload, ...) can crash the store several times in quick succession. With OTP's stock supervisor budget that's enough to take Snippy down entirely.
The default budget here is forgiving: it tolerates roughly a 20% failure rate on a server handling 50 requests per second over a 15-second window (about 150 failures). Tune via application config if you want tighter or looser bounds:
# config/runtime.exs
config :snippy,
max_restarts: 150,
max_seconds: 15