I get your point but the argument posted is a bit specious. I don't really want to "get into it" on this forum, but the arguing that because not everything can be included, that nothing can be done is a bit obtuse. It's hardly a practical deduction for life, i.e. there are too many roads to Los Angeles, so let's never drive there, or, there are too many possible letter combinations, so let's stop using words or making dictionaries. There are also far more extensive projects than drivers for the ESP8266 chip in it's various packaging and versions that do survive quite well in revision control repositories. NetBSD, for example, supports 8 Tier 1 cpu architectures and 49 Tier 2 architectures. FreeBSD's (and MacPorts/Darwin ports) port system has over 12,000 third party applications that can be installed as binaries or compiled. These are projects with millions of lines of code in a variety of languages.
My point, if it isn't yet clear, is that not everything would have to be included, but one could choose areas where the benefit for the company is the greatest and the chances of completion are reasonably good. The problem is not monolithic. It can be subdivided into manageable pieces and these can be prioritized.
Since I'm probably not going to do the work, this is only a suggestion. I just found the reasoning posted about why one shouldn't do it not very logical or helpful.
S.