How to implement rate-limiting for Confluence Cloud?

Hello, me again, please can someone confirm how we’re supposed to implement rate limiting for Confluence Cloud?
The docs don’t mention the “cost budgets” that Jira Cloud has, so should we consider all requests made from an app as being part of a single “cost budget” to steal Jira’s terminology - ie if a single request to Confluence Cloud receives a 429 or other non-200 with a rate limiting header, then we should delay all further requests to Confluence Cloud for all endpoints from all parallel processes/threads/microservices until the relevant time period has passed?