BPMN and BPMN-based or related tooling is pretty popular within businesses for, well, business process automation. It offers a no/low-code way of moving manual processes to an automated form and can work well. But these are relatively simple programs that often boil down to a dataflow-style + decision tables. The diagrams are desirable for communicating the process to non-programmers anyways so it's handy that you can (more or less) use them to at least describe the flow of the system, if not all actions (like you may write code for some of the actions, but overall you can accomplish most things with the diagrams and tables and built-in tools for your particular BPMN service).

I had similar experiences with LabVIEW 12+ years ago, it was effective for what we were using it for which, similarly, boiled down to a dataflow-style program and decision tables pulling from various data sources and pushing to various data sinks.

The biggest challenge I faced working on a team with things like this was that scaling to more complex systems only works if your design is reasonably modular. If it's not, then it's hard to simplify the current presentation sufficiently to make it comprehensible above a certain size of diagram. Of course text-based programs have the same problem (the 10k SLOC function with the same 100 lines repeated throughout but in slightly different orderings and with different variable names), but most programmers are more familiar with that situation and can either slog through it or know how to refactor it to get out of the mess.