A lot of IRC usage used to be purely social. So the fact that that are lots less users isn't really indicative of anything in the context of FOSS development.

Indeed, in the same time period I think Freenode actually grew.

Good point, usage for gaming has also died completely. Old networks like QuakeNet existed to allow gamers to co-ordinate match making. Now who needs that?.

IRC population has also become harder to track because more people now set up cottage networks that have one server and only host a few channels. There was an article a few years ago discussing this exodus from big networks which I'm having trouble bringing up now.

The worst that can happen to IRC is something like Usenet's death - growing too popular to scale. But scaling isn't of primary importance because a network hosting 100,000 channels is only mildly more convenient than 100 networks hosting 1,000 channels each. Nobody wants a chat stream that looks like a popular Twitch channel, because there is no longer any signal when that happens, so big channels are outliers. (And Twitch chat is IRC-based itself - demonstrating how much this technology is a commodity.)

IRC isn't going to stagnate completely until people decide to give up on text-based chat. The main thing it has trouble with is new client features like inline images.

IRC's userbase is consolidating around the OSS crowd, hence Freenode exploding in popularity while other networks falter. Unlike regular consumers, OSS people understand the value of minimum-common-denominator networks, interoperability and openness.

When most people will do all their chatting in Facebook or Slack, OSS developers will still use IRC.

Things decreasing doesn't necessary tell anything worthy. IRC is less used, but it's still used, and used a lot, and guess what, one can find a lot and lot of very rare and interesting people and data there. All of this without heavy hardware nor software requirements, or even psychological, it's still the same old IRC system, no change required.