async task queues rabbitmq vs kafka vs sqs

hey everyone, i'm wrestling with choosing the right async task queue for a new microservice architecture. we've got background jobs like image resizing, email sending, and data processing that need reliable execution without blocking user requests. rabbitmq seems solid for traditional queuing, kafka is appealing for its stream processing capabilities, and sqs is the aws-native option. what are your experiences with these? when did you choose one over the others, and what were the trade-offs? looking for real-world pros and cons to guide my decision.