From lab to libre software: how can academic software research become open source?

Its not all doom and gloom, though i get where you are coming from. In astronomy things are changing slowly, there a several big projects that get funded for software "development" (with the science side almost being a secondary component, atleast when writing the grant) have a look at the NSF's SI grants which fund software infrastructure. Also the journals are slowly changing, ApJ (a top tier journal for astronomy) now accept code papers that just describe the code without needing to do new "physics" with them (see http://journals.aas.org/policy/software.html). In fact i work on a open source astronomy code, with several software papers, and because it is open and other people get to use the code, these code papers get a lot of citations and have become my most highly cited papers.

The other useful thing that is changing as well is citing smaller projects, i maintain a plotting library that on its own is never getting a paper, but after doing a release on github it gets picked up by zenodo (https://zenodo.org/) and gets a a DOI so i can cite it in my papers. ApJ now have a section you can add to your paper mentioning (and citing) the software you use, including ancillary things like pyton/numpy/scipy etc, not just the "main" code you used.

Though i do still find when interviewing for jobs people question the value of my software development (even if i have fixed bugs for them and they depend on the code). Maybe software development needs to be seen more like instrument builders, something that is essential and should be rewarded but not necessarily going to generate the huge papers on their own.