That applies if your only goal is to create FOSS software that competes head-on with proprietary applications, so that being FOSS is the only differentiator.
But obviously, this isn't always applicable as a general principle. If you want to enable new functionality and support new use cases, where there are a variety of approaches or trade-offs, and conventions aren't well-established yet, you probably shouldn't presumptively conform to an a priori monoculture.
> Less than pleasant because there's a pretty significant amount of overlap, but none of it is able to be homogeneous as there is some hard coding that a keymap alone doesn't allow for convergence.
This is a great counter-example, where adhering to well-established conventions for problems that are already solved is advisable. There's been a lot of divergence in UI design recently, with lots of applications moving away from decades-old standards for purely aesthetic reasons, and creating huge usability regressions in the process.
Hypothetically if GIMP just copied the photoshop UI and expected behavior a lot of tutorials for PS translate directly making GIMP accessible, which ultimately plays to the favor of GIMP as it is more accessible in that it's open source, free, and doesn't require subscription or accounting.
Also since GIMP is decentralized and volunteer-ran it'd be interesting to see exactly what Adobe would do about it, if they could even do anything.
Something else that would be nice to see from the open source community would be standardization in the control schema across software. For instance I use Blender a lot, but I also use Inkscape, and jumping between the two is... Less than pleasant because there's a pretty significant amount of overlap, but none of it is able to be homogeneous as there is some hard coding that a keymap alone doesn't allow for convergence.