Maximum restart frequency limits how many times a supervisor may restart child processes during a configured interval. The default permits three restarts within five seconds. Configure the limit with :max_restarts and :max_seconds passed to Supervisor.start_link/2. section 8.3.7
The limit prevents a supervisor from retrying an unrecoverable failure forever. When restarts exceed the threshold, the supervisor gives up, terminates itself, and takes down its children. section 8.3.7
Exceeding the default
Start the system, repeatedly kill the to-do cache process, and pause briefly between attempts. Four restarts within five seconds exceed the default, causing the supervisor and its children to terminate.
Restart frequency also shapes supervision-tree design. Separating database workers under their own supervisor confines repeated restart failures to database operations, allowing a parent supervisor to restart that service without disturbing unrelated processes. section 9.1.5
