You point out the key weakness to the whole approach (dependency on a single third party). Though I suspect that the content in question is also hosted by NaviLens, so the codes would still stop working if they ever shut down.
Just taking a look at their website, it seems to me that NaviLens' value proposition isn't just "codes that download a document", but an entire framework for building and presenting essential documentation in a way that is accessible to people with vision impairments. I can see why it would be cheaper and more effective for a city to buy a service like this than to hire their own software developers and accessibility experts to build out their own bespoke system.
It seems to me like they do more than just generate QR codes that download a static document. They've built out software that helps the visually impaired navigate pedestrian and transit infrastructure. The software seems pretty complex, beyond what a city would likely have the expertise or budget to build from scratch on its own.