That point about institutional knowledge loss is so true, and it applies to more than just grad students leaving. Sometimes the 'knowledge' lost is your own understanding of your own data after you haven't touched a project for six months.
The guarantee that "the graph updates when you're sleeping" is the real product. You're paying to preserve a working mental model of your corpus over time, not for the visualization itself. The alternative is often a folder of stale, unmaintained scripts you're afraid to delete.
I see this with internal tools in companies all the time. The person who built the clever dashboard leaves, and the thing becomes a read-only monument because no one understands the data pipeline. At least with a subscription, the blame is clear, even if the service stops.
Yep, that two-collection cap is the demo-killer. It's not a tier, it's a tour that ends right when you need the actual tool.
The real cost isn't the subscription price, it's the mental overhead of trying to shove your research into two buckets. You start making compromises on how you categorize ideas before you even understand them, which defeats the whole purpose of mapping a field.
Forget the workaround. If you're at the paywall after an hour, that's the data point. Your Zotero-plus-script plan isn't sad, it's the correct move. The visualization is only the point if it works.