when do you add a message queue between services?

hey everyone, i'm wrestling with a common architectural decision and wanted to get your take. we have a growing microservices setup, and direct synchronous calls between services are starting to feel brittle. failures in one service cascade easily. i've been thinking about introducing a message queue (like RabbitMQ or Kafka) to decouple services, making them more resilient and scalable. but when is the right time to make that jump? is it when we hit X number of services, Y number of requests per second, or when we start experiencing specific types of failure patterns? curious to hear your experiences and criteria for adopting asynchronous communication