Nonlinear retries

[jelder]

Errors are retried immediately and forever. Would be good to enter a truncated exponential backoff loop or similar. A full jitter algorithm is available and in use by the AWS client. We should make sure failed messages are retried after an interval derived from that.

sqs-lambda-bridge/src/aws.js

Lines 4 to 15 in eba837a

	/*
	https://www.awsarchitectureblog.com/2015/03/backoff.html
	Explore in the console with:
	[1,2,3,4,5,6,7,8,9,10].map(fullJitter)
	*/
	const cap = 30 * 1000
	const base = 100
	const randomBetween = (min, max) => Math.random() * (max - min) + min
	const fullJitter = retryCount => {
	const temp = Math.min(cap, Math.pow(base * 2, retryCount))
	return temp / 2 + randomBetween(0, temp / 2)
	}