I wonder how many 1000s of hours of some of the brightest minds in science are being wasted creating cool #rstats packages that only 10 people will ever use to shave 5 minutes off the time taken to solve a problem and which will stop working with dependency issues 2 years later.
I don't follow all that many people and yet every other day there's a new cool package in my timeline that only a few people will ever have time to learn. But someone will use it, and in a couple of years time someone else will want to run their code, and it will choke.
The standard answer to this seems to be "But it's open source, so you can fix the package when it breaks". Great. So now anyone who wants to use code written by A has to be able to fix B's package, of arbitrary complexity.
I don't have any magical solutions for this, other than suggesting that people think about as long and hard before producing new packages as they would about buying a dog.
