I think illustration isn't something too much in the mindshare of open source, so overall support for it isn't great. IMO this has contributed to it. The industry standard tools are all closed, with closed formats, so it just sounds like much more of a hassle vs contributing code/text.

I mean this throughout the whole process. The only standard illustration file format I can think of is SVG, but it's largely a format to export to, not one industry-standard software uses as it's main persistence format.

So for starters, contributors tend to need access to speciality software they probably don't have installed to view and edit the source of truth. This also means you're handling at least two files in your VCS, the closed format acting effectively as a blob, no diffs, etc. and an export file (usually more, for different scales) to actually interface with the rest of the ecosystem; this is the file everyone can open, inspect and compare, the one your build consumes, etc.

This already would be a good amount of friction for someone familiar with the tools, but designers are not necessarily familiar with git, the PR process, etc. Add to it that icons are more subjective than code, which overall should follow certain rules and either works or doesn't, and it overall seems not worth it for a casual contributor.