My option is that the site life cycles are somewhat broken. Communities are too and need help.
Originally there was a focus on growing these smaller communities - AKA Stack Exchange 2.0 and the associated projects, such as team CHAOS, and having dedicated, fairly engaged staff across the company on many of these sites.
Unfortunately the choices that were made -– a rather strong focus on SO adjacent products – ended up meaning most of the smaller sites got neglected for years.
In addition, Communities have hashad a lot of damage done to them over time by choices made by the folks who run the company over time. Healthy communities need commitment, care and work. In addition to neglect, we've lost/had people driven away who were good at running these communities.
While to an extent communities have been resilient the sum total of this is the core of many communities have gotten gutted, and fixing this hasn't been a priority. And this means outside a few hardcore people remaining, a lot of communities have not recovered from traumatic events, or people have simply drifted away.
I'd say (optimistically) it'll take a few years of some degree of dedication - nodedication–no getting distracted by the next shiny object, and downsizing of community centric staff to make profit margins look nicer to make a dent on it).