Mats does not explicitly define a set "route" through a bunch of stages. I view Camel more like a Workflow system, which is pointed out elsewhere in the discussion threads here - the workflows are defined external to the actual processing steps.

With Mats, you define Endpoints, which can be generic like a REST endpoint can be generic: "AccountService.getAccountListForCustomer" - which can be used (by invocation) by sevaral other endpoints.

Also, IIRC, the process you show with Camel is now effectively thread bound - or at least JVM bound. If the node running that process goes down, it takes all unfinished, mid-flow, processes with it. The steps are not transactional.

With Mats, the "life" of the process lives in the message. You can literally bring down your entire system from GCP, and rebuild it on Azure, and then when you start it up again, the mid-flow processes will just continue and finish as if nothing happened - as long as you brought along the state from the message broker (in addition to your databases). Viewed "from within the process", the only thing visible is that there was a bit more latency between one step and the next than usual. AFAIR, you cannot get anything like this with Camel. The idea is of course not the ability to move clouds, but that the result is exceptionally robust and stable.