Retry Budgets will not help stopping retry storm?

[rafatbiin]

According to this document it says that retry budget will help stop the retry storm. but I don't think so. let me explain why.

Imagine we've three services A -> B -> C and A is receiving a steady 100rps. we define the retry budget at 20% from A to B. in this case we can have a budget of making 20 retry-able requests. which is fine from A to B but now B will have 120rps of active requests and let's say it also configures its retry budget for B to C at 20%. now, we'll just increase our retry budget on the downstream C because we're having more active requests on B. as we move down to the path, this retry budget will keep on increasing thus creating a retry storm.

please feel free to correct me if my assumption is wrong and please suggest if you have an idea about how to stop retry storm in a mesh. if you face similar issues on prod, I'm interested to know how you solve them. Thank you.