I would love to hear more about why you think Event Bridge solves the Lambda Pinball and when and why it is a better choice. I would have thought that poor architecture without Event Bridge will still be poor with it.

The thing that really puts me off at the moment are
1) More complex integration than SQS/SNS/Kinesis
2) 300 events per second limit

The 300 events per second limit is concerning given that you could have 3000 lambdas executing concurrently with 100ms response times. For bursty loads I can easily see potential for throughput limits being exceeded.

On the plus side I really like the Event Registry and the ability to integrate with third party services (although I’ve never needed to do this in a way a webhook doesn’t solve).