I think that it will be great if we can increase the sleeping time between each retry.
So in cases where there are failures and the retry mechanism begin to work , we should increase the timeout between the retries rather then try again (if there aren't any othe notifications to process) after 250 mili sec for each retry. That way if there were failures due to network connectivity for example, it might succeed in the next retries.
Definitely not a bad idea!
I don't think this request is relevant anymore given the way 2.0 will handle notifications. Basically, if a notification fails due to a network problem, it will not count towards its maximum number of retry attempts. Each platform is also responsible for its own exponential backoff or maximum connection attempts from attempting to connect and having network failures.