Also, lag issues are not really limited to high-end graphics / video editing. Sure, lag/low performance makes those tasks basically not doable in the browser. But even "ordinary" webapps - like word processors or issue trackers - are annoyingly slow. Take Jira for example - I'm probably wasting a few (billable) hours every month just because it's not a desktop app. So are all of my co-workers. It's most likely also the reason people tend to batch time reporting and do it once a week or so (it's much less distracting then), which as a result precludes management from having access to up-to-date information about time spent on tasks.
Trivial inconveniences can have non-obvious impact.
I think you'll find that Desktop apps often have to communicate with with centralized servers as well.
From what I've heard (no in depth knowledge so I welcome corrections), the DOM / CSS model itself really wasn't designed for fast, responsive applications and thus tends to under-perform in that department. So even a fully internal network app with comparable communication speed may seem slower on the browser from the get go.
Many web apps often also tend to be a little bit laggy just based on their cloud based nature. Load a random spreadsheet on Google docs, and you are probably talking 100-200 requests and (depending on your network) perhaps a 5-7 second delay. Change a tab, and you probably have another 2-4 second delay. Compared to the 1 second load time of a spreadsheet and nearly instantaneous tab switching time on a native spreadsheet app, it feels slower... especially for the power users, I'm sure...
Then again, from what I know, Premiere, Photoshop, and DAWs like Cubase and Logic are straight-up C++ applications, and don't even go into Java / Qt territory.
I would say that something like Google Apps vs Microsoft Office is a lot closer than it used to be. Native-app Excel sheets for instance still have the edge in speed overall, especially for large datasets (say, >10,000 rows) better, and Excel has a pretty significant edge in my opinion for analysis (pivot tables, charts, etc.). Google Sheets however are very capable for a large percentage of documents, with the added benefit of being so much easier to share and collaborate. Neither use case is a "toy" scenario in my opinion.