Minor: I think it's good to discuss the issues you're interested in from your area of expertise with neuroscience. You even state outright that you'll be framing things from that perspective. However, the both the intro and that outright statement suggest the paper is written for a more general audience. That conflicts somewhat with the framing here, which seems to be geared towards neuroscientists familiar with the features of neurodata. Not a major concern though, this is just a question of whether you want to maintain a consistent audience throughout the paper. Arguably reconciled by substituting "We" with "neuroscientists". Same comment applies to other areas where this may pop up

Minor note, but the framing "beneath the level" may be counter to the overall argument I'm getting, which is that each component that results in the publication should be considered significant. However, I can also understand wanting to retain this language for an audience more skeptical of that argument, which changes the framing to "these components should be valued to some degree and we don't currently reward that"

May be interesting to look at researchers that contract other companies to perform this type of technical work for them. For example, Aptima produces a variety of tools to assist with measurement of team dynamics. Researchers that contract Aptima to use these tools obviously recognize and value the development of these customized tools, which suggests an awareness that this type of technical knowledge should be credited somehow.

And even then, the use of a tool like Slack basically means the information being produced is held hostage by a 3rd party company. Access is granted in exchange for payment, even if the access is not particularly worth paying for due to the poor organization embedded in any chat tool.

Interestingly, I think this phenomenon is directly tied to how every scientist is expected to keep up to date with the latest standard of evidence in the form of new statistical and analytical methods. Programming is the most efficient way to incorporate and teach the rapid development of these methods, so sometimes that just results in people being a "programmer" by loading up a package in R and calling up the relevant function, regardless of whether they necessarily understand the purpose of the tool. Of course, some tools are easier to misuse than others. I suspect that computational modeling in general is a lot more like traditional programming than other types of statistical analyses.