The thing is, you solved a very narrow problem to which there is already a solution (and has been) on the mainframe for 30+ years (MQ). The real problem is that in that COBOL code is 50 years of business rules smeared across millions of lines of code, adjusted for all the changes in law (sometimes applied retroactively) which impact how money is handled. It isn't that mainframes don't have message queues or can't interoperate with web services (they can), even if not all customers take advantage of those features. The problem is replacing that code requires extracting all that knowledge out of the code. Then, on top of that, if there's any downtime, it can be existential risk for the bank.
Yep. It's real easy as an architect to come in and propose an overall architecture that will work, but the rubber meets the road when you attempt to 'strangle pattern' your way out only to find the deep interconnected and undocumented business logic. You're also fighting the business by trying to wrangle SMEs that have no interest in helping or have long since moved on.
