like the title says its my opinion

a lot of time can be spend on it ;)

intresting information about it... Thanks for posting valuable content. your content helps me for it.

Definitely! It is a really interesting and a fun process to go through after dealing with stuff like coding, bug fixing, level design. Also I sometimes see videos about making UI elements like inventory and it makes me really excited for it as well.

More importantly, it's a role on its own, even UI programmer is a role

Why? Because it's so damn important and complicated. Even if you make a good game and skrew the UI, then you skrew the game as well: the players won't be engaged and/or won't understand how to play, what to do

I spend months and months, at this point even more than a year to improve it, and it never seems enough

I agree albeit in my opinion UI design is mostly a craft you can learn and a small part an art where you rely on your talent. I have approached several projects with making a matrix (Excel-Sheet). The rows are the list of UI-screen. The columns are the criteria I consider necessary for a good UI:

• Usual UI conventions (Mouse for look, WASD for movement etc.)
• Appropriate aesthetics for the game setting (e.g. no baroque font for a Scifi game)
• Same colors, fonts, button styles, and icons across every screen
• Minimalist design, as little text as possible (except for credits and story)
• Visual hierarchy (e.g. Play button bigger and more central than e.g. audio volume)
• Readability including different screen sizes
• Immediate feedback (wait cursor if necessary)
• Screen shows where I am (e.g. by headline/picture/background)
• Screen easy to find/open and easy to exit

In the matrix I mark every box whether I think the criterium in the column is met for the specific screen (I use a plus for good and a minus for bad) and fix what's bad.

Those criteria don't replace play-testing. It's more like a check if the UI is ready for play-testing. Typically I'd do some fine-tuning  according to the feedback from play-testing.