I work in psychiatry research, analyzing and maintaining the sexy fMRI neuroimaging data. I also write the storage and analysis database that we use. The database usage has been growing exponentially as data sharing projects have started and the NIH has mandated data sharing. In other words, my workload of maintaining this software system has also grown exponentially. What my PIs do not understand is that software is not at all like scientific papers. Once one of their analysts (or post-docs) writes a paper and gets its past reviewers, its done. If there is a major or minor flaw, chances are good that no one will notice or say anything.
It's completely different with software engineering. If there is a tiny bug, people will notice. Having transitioned from analyst to programmer, my work is viewed entirely differently. If the papers published from workplace underwent the same scrutiny that the software does, we would produce much more robust science.