You can have all sorts of fun in "Advanced Options > Advanced Rendering Options." The relevant settings for these are listed under "Rendering control." I don't believe you can disable mipmaps entirely, but using either Nearest Neighbor or Trilinear (according to preference) with 16x anisotropy (no point in going higher) will get rid of the obvious detail lines.

Note that nearest neighbor also affects things like lights and shadows, which might look pretty strange. The game is built with filtering in mind.

It's not quite a pistol start; the game gives you a set of default weapons on a per-level basis. I wouldn't call a fresh start something almost standardized like it is in the Doom community, though. The game is clearly intended for continuous play. However, it is also designed around the default weapons being entirely practical. In the end, it is a preference thing. I like the slight challenge boost.

For a first time ever playthrough, definitely choose Hard. Serious has double damage, which is real harsh if you aren't familiar with the general landscape and enemy behavior. Serious is well worth an eventual revisit though, since the extra enemies (and double ammo!) wildly change how the game plays out.

My other general advice is never quicksave mid-fight. Often you checkmate yourself long before you die, and it's really easy to save where, it may not look like it, but you've already lost.

When I play classic Serious Sam, I use this port with Vulkan rendering. You can just grab the AppImage and drop it alongside the original executable. I believe it also fixes the infamous half fire rate bug. (I'm genuinely not sure; I usually play each level "from scratch" via the level select, so it would never be a problem for me in any case.)

However, the original game is still perfectly playable with a simple config edit for higher resolution and FOV.

Here's one I know. Creates aesthetic wavy things.

https://tanck.nl/wallpaper/

It is an HTML/JS pair, so it would require some wrangling to get it to write to a file instead of an HTML canvas. Though it looks like someone translated it to Python with file output here. (I have not tested this.)