I'd argue that such information consumers are largely the same audience that wouldn't want to muck around with an API or remote service to begin with and aren't the target audience. When I check the forecast, I just use one of probably thousands of services that regurgitate and slightly alter NWS data. Just because I unlike most data consumers could read the data a different, more arcane way, doesn't mean I would (I absolutely wouldn't, I'm going to be lazy).

On the other hand, when I have been interested in lower level access to data and not just data products, the time I'm looking for data repositories and APIs to work with the data at a more detailed level, the time these sorts of services come into play, I'm very often (I'd say the majority of the time) quite interested in historical data. I want to look for trends, I want to look at how accurate forecast models were, I want to compare historic radar with forecast with a ground measurement station. For some I imagine they want to throw ML and other AI related techniques at it. Those are the exact times these sorts of services are useful to me, in my eyes.

So I agree with the obvious premise that most consumers of weather data just want to see immediate forecasts that effect their future, they care nothing about the past, I think those aren't the typical users of these types of services and were looking at two different user base targets.

Because there are many use cases for historical data (algorithms' training or its cross-analysis with other data, for instance), while lots of APIs focus on forecasting. Maybe at this point we need more innovation in either the forecasting itself (generally "better" forecasts, especially for industries, and better extreme conditions alerting), or innovative use of historical weather data in other domains.