Sure, but that's just moving most of the chaos from your app to a message bus. Which is almost always correct, assuming you are actually at the scale where you need that level of architectural complexity.
I wouldn't agree that it is moving reliability issues to the message bus. This is more a scaling/availability problem than a reliability problem, but the message bus is definitely helping to solve scaling/availability problems.
While Kafka can indeed be pretty complex, streaming architectures aren't necessarily complex at all. For example, having a AWS Kinesis stream sitting in front of an AWS Lambda is pretty simple, and provides efficiency benefits even at comparatively small scale.
