For people that still develop desktop apps, do you see any signs of more people leaving this line of work? In the long run I only see them remaining relevant on industry-specific work like CAD or scientific computing. General office work doesn't equate to desktops either and even less to desktop applications. Most everything is in "the cloud" (why use Office when you have O365). Do you find the dying practicality of desktop apps to be accurate, or overblown?
I think the death of desktop apps to be a bit of an overblown claim. But it's true that developing a fast, native desktop app can often be more complicated that building an online CRUD-style app.
Some applications have enjoyed success by moving from the desktop to the cloud, although in cut-down limited form (think word processing, spreadsheets, presentation software). And yes, there are visual design tools too (e.g. Figma). But a lot can't be done in the "cloud" and cloud apps offer only a subset of what a desktop app can do. (Sometimes a subset is all you need.)
Perhaps the future is more hybrid apps: desktop apps with some cloud capabilities such as online backup/saving/sharing. I can't imagine heavyweight vector, bitmap, video, or animation tools being anything other than desktop apps for the foreseeable future.
And don't forget that some cloud services offer desktop clients for connecting to their online services. That suggests that there's appetite and plenty of life left for the desktop. How many developers would give up their desktop text editor or IDE for a cloud-based one?
And finally a personal preference: between the choice of an online app (that is more than just a simple CRUD app) and a native desktop app, I would pick the desktop app every single time.
And I think you can add a lot of heavyweight 3D software (Blender, Maya, ZBrush), game engines (UE4), and audio software (ProTools) to that list as well.
On the developer end, what would you do if you had an EC2 C5XL with a massive disk for each and every one of your users, for free? From a startup scenario, there's runway to be gained by offloading storage and processing to the client, desktop can bring monthly ops costs down to the basement.
On the consumer end, a single user perpetual license can be extremely cost effective. I'm watching the Google Stadia release for an idea of what a resource intensive cloud solution will really cost. I think it remains to be seen.
Desktop is certainly becoming niche--but I think it will rule certain software domains, specifically those that have high storage and processing requirements, for a long time.
The Windows code signing certificate I have is an OV certificate which means it builds up trust over time before the installation security warning goes away. It seemed to go away within a week of release, but was still frustrating to deal with. An EV certificate skips the trust building step and can be purchased for about 5x the cost.
A couple months ago I switched to a Progressive Web Application (PWA) which reduced costs and removed the installation requirement. There are capabilities lacking in web applications, but there are planned features such as native file system API [1]. Hopefully these also make their way into Firefox and others. No desktop PWA support in Firefox at all yet.
[1] https://developers.google.com/web/updates/capabilities
1. Increased data privacy issues as more and more companies either sell information or get hacked
2. Ongoing cloud costs, which sometimes increase in large amounts as companies restructure prices and feature sets.
If there are 3 competing products, and one offers a Desktop application in addition to their cloud offering, then I think quite a few customers will choose the Desktop version due to the 2 points above.