Obviously we can machine generate stillborn languages as quick as we have time and money for. The question at hand is, how many of these can be spoken? Almost by definition, the auxlang and it’s fans can tolerate only a few languages, maybe one or two. If a language is to be a language for speaking to random people, it needs market share.
Natural language can get by on a few dozen speakers, just so long as the community is isolated, doesn’t mix too much with the surrounding world and so on. By this measure, the world could support this many languages: Total Population/a few dozen. And that’s a lot of languages. In practice, people don’t stay in isolated communities, the economic costs are too high. They move to larger communities and the more widely spoken the community’s language, the more attractive it is, economically speaking.
For the world of conlangs, philosophers, mathematicians and logicians were an important audience. Pop culture fans another new important audience. These audiences are not concentrated, so it hasn’t been until the internet that these could be language communities instead of socially inert reference grammars.
If natural languages need to have a geographically isolated and stable community of speakers to survive, then conlangs, I suppose could learn from that. For example an intracity or intraneighborhood language, if it could attract the intended audience.