I think I made a mistake by saying "information is lost", rather than "some information is lost" - clearly, some information remains, and as I started my comment, I'm not saying it's impossible.
In fact I share the sentiment that seemingly irreversible things can often be reversed, we just don't have the technology to do so. However my enthusiasm took a huge hit once I learned how the technology we have usually operates: for example the photo sharpeners enhance photos by creating faces different than the actual original pre-blur faces. Just got an idea I have to test eventually, how close would those "recovered" images be to the algorithm input, if the blur was reapplied...
Your first example (recovering text from photo) is impressive, but the blur is clearly one that blurs as little as possible to make the text unreadable. I think if you increased the blur just slightly, too much information would get lost. I also wonder what was the resolution of the image and what was the color depth (was that a JPEG or RAW format?) Without the ability to inspect and repeat the process I'll remain skeptical about de-blurring.