And here's the thing: I understand, at a very very deep level, precisely why my application is slow to start up. I understand it at every level from the semiconductor gates in silicon up to the color specification lists used by the GUI toolkit. But none of that helps me make it as fast as the handmade community thinks things should be.
Why? Because I know that the slow startup is caused by substantial amounts of code that all add useful functionality for at least some of my users. Do I want to get rid of i18n? Shall I skip checking for new versions? How about forcing plugin scans to be initiated by users and never doing them automatically (don't even think of suggesting doing them in another thread, and don't ask me why)? Shall I skip the ability to handle RTL text? Should we drop human readable and editable configuration files in favor of some faster and less resource-eating format? How about a scalable GUI, is that really useful for most? How about a themeable GUI, given that it's not actually that themeable anyway?
And so on and so forth. Each one of the above (and lots more) contributes just a little bit to the slow startup. For most users, all but one of them will likely seem not very important. So which ones do we drop to get back to the sort of startup times that the handmade community seems to expect?
Because let's face it, xterm can put a window on the screen before my finger is off the return key. So why not a DAW (or an IDE or an editor or a vector graphics app or an image editor or a browser or a circuit emulator or whatever your thing is) ? Just which features do you want to cut to accomplish this?