lucynoodles20

[crystal aspen]

Hi 👋 I'm not sure off-hand if we identify in the response is a replay of a previous request, but wouldn't the retry logic you implement to make those idempotent requests already have this insight as it is the code triggering the retry?

[olive dock]

We are not currently using idempotency keys and are getting double charges for a number of reasons including (1) race conditions between manual and automatic retries and (2) "blips" on Stripe's side, where the initial response was blank or an error, but actually succeeded. We want to add idempotency keys but are concerned that this may accidentally prevent some legitimate retry attempts. Since these are some weird edge cases we're dealing with, it would be nice if there was a clear marker on Stripe's end to say whether we're hitting an idempotency key.

[crystal aspen]

I'll see if we do anything like that, but I'm still not grasping the benefit of that. If your retry logic knows it is using an idempotency key, then it knows that already, if it doesn't then it doesn't seem like it would matter whether it's a replay as you need the response anyway.