This issue comes from this issue #42 (comment)

As a rule of thumb, in all code that works with the network, the network will always one day fail.
For that reason, it will be great to implement a retry logic inside of the action so that the requests made to the GitHub API have a retry logic.

Also, we can expose the retry parameters to the user.

  attempt_limit:
    description: Number of attempts
    required: false
    default: 2
  attempt_delay:
    description: A delay between attempts in ms
    required: false
    default: 0