I whole-heartedly agree, but the reality is that it depends on your perspective. We could have uni kernel apps to run spreadsheets, but instead we now use Google Sheets, inside a JS VM inside of a sandboxed browser on top of an OS which maintains a separate copy of tons of configuration files for each separate piece of software which runs in a sandboxed environment and uses other layers of translation to do things on the hardware.
This state of affairs is utterly ridiculous and bad engineering. Sure it's driven by the market, and most of us have to operate within those constraints because of our jobs - but I think it is a responsibility of a developer to push back at least a little bit when the market pressures promote bad engineering. Think of it as sending feedback to the system. That's the only way to price quality in - users won't do it, because in most cases, users have zero real choice except to use the product. Only those making it can make it work well.
And then we can also support those working on changing the platform-level stuff.
> We've shown over the past 40 years that we value developer productivity much more than user resources
Of course: we're developers, and we can't multiply. When we use the user's time, we picture using one user, not one million (or, more reasonably, a couple thousands). So it rarely occurs to us that the users' resources are vastly more important than our own productivity.
To a lesser extent, this can be true for custom software as well.
We've shown over the past 40 years that we value developer productivity much more than user resources. Most relevantly, consider the average size of a webpage: http://www.yottaa.com/company/blog/application-optimization/...