In practice, "what's actually best for the community" (growing a community of engaged engineers who feel listened to and who understand what you're doing) is not necessarily easy to show by metrics. I think the author is making a subtler point that, even if you don't hate DevRel work, the most useful kind of work is often unappreciated or devalued by decision makers.

True, but in practice "the community" as the DevRel person sees it can be very different than the customer base that the company wants to address.

For example, this DevRel person talks a lot about their Twitch stream and all of the work they put into a game on it that also used Sentry (their DevRel employer). It's all very cool and impressive work, but for a company like Sentry with 100,000+ customers there are millions of engineers across those companies that make up their developer base. Streaming to a couple hundred people on Twitch doesn't even register as interacting with the community.

Reading some of the linked posts there's an ongoing frustration with being asked to provide metrics and prove value, which gets to the core of what I was trying to say above: What some people want to do for DevRel for a small audience (Twitch streaming, podcasts, workshops) is often at odds with what the company expects them to do for their large base of active developers and users.