My take is that there are two factors at play here:
* Third party requests from the browser are problematic in general for technical and privacy reasons: it's impossible for the browser to know whether the request is privacy-preserving or not. And they need to be enabled in tooling like CSP. It's much easier not to have any.
* Third-party requests from my server don't give me any more information than I had before. I could store my logs and process them, or I can engage a third-party to aggregate my data for me. Plausible have no way to know that the requests I'm sending aren't entirely fictitious, and I can sleep slightly more soundly knowing that there's one less moving part that I need to maintain myself.