I agree with all that. I also did a PhD in fluid dynamics, and I have programs that I used for my thesis results, and that could be made available and might one day prove useful to someone else.
But they're very specific, combine functionality in strange ways - e.g. mixing computation and display - and partly duplicate function that I'm sure has since been done better in other free software programs.
Hence the upshot is that it's really difficult, even for me as a person with some track record in free software, to see how this code could usefully be made available.
The question of reproducibility of research, and whether journals ought therefore to require associated code to be published, seems to me to be separate from the general desire to share code. But if the former was widely required, it would probably also facilitate the latter.