A lot of single scenarios have a goal to find artifact.

Some objects on a map (crypt, treasure chest, wagon, warrior's tomb, utopias, ship, shipwreck, black market, sea chest, shipwreck survivor, skeleton etc.) may give you artifact with some probability, and if you are lucky enough, you can finish mission immediately.

We may call it "Artifact rolling" and I find this a very bad idea that break sense of speedrunning for some missions, and I vote for banning such technique. I encountered this during first LLtQ mission, when you need about 300-400 rolls to get Boots from Shipwreck Surviror and thus I can save about 2 days playing In-Game Days category. I consider this as a glitch (glitch is a thing that changes gameplay significantly) and didn't use it.

HOMM3 is an amazing game which may have a lot of improvements, but if the only improvement you need is to roll artifact 1000 times, it became boring.

Yes, some of you may say the same about Hero rolling (Invite Hero for HD-mod) and Spell rolling. But strong hero do not skip half of mission. And Spell rolling is officially acceptable by community (use of Guild Mage rebuild). So these both are fine for me.

To be honest, I am a bit torn on this issue. Artifact rolling is indeed an entirely luck-based procedure, but it can vary drastically in odds:

SoD Grail-rolling (which could be assigned to artifact rolling) is 1/25, so better than hero rolling and secondary rolling at times

AM/BM variety of art-rolling (using this term from now on) in EoL is 1/38. Bad, but not ridiculous. Minor artifacts rolled in them have even better odds.

Shipwreck survivor is bad Treasure Chest is ridiculous

The most infamous is probably rolling crypts for Armageddon's Blade in FFA scenario.

Personally, if it ever became a serious issue, I would opt to make 2 categories: Any% and Any% No art-rolling, at least for IGD leaderboards.

For IGT, I am not sure what to do, but the problem is so far removed from the current state of leaderboards, we might as well never live long enough to see it become a major concern. All IGT runs are improveable without the use of art-rolling, maybe with the notable exception of EoL, but then again - that is not even close to the ridiculous odds of treasure chests. One solution I could see here is drawing the line between AM/BM art-rolling and the shipwreck survivor kinds.

If players find a way to better know or predict generation rolls they may even fork or copy whole game process (at least in linux system) to reproduce game random as easy as well. And resume playing with 0.00 on timer on any comparable machine. *sorry, idk about current timer start rules and cathegories

Lets see it as a preparation phase for the run just to reduce ridiculous preparation time to remake needed starting artifacts or other staff. It just normalization.

As far I am aware (from my own testing), you can, to some extent influence random aspects like secondary skill rolls or random creatures & followers (they are random, but not independent, so you tend to see some "patterns" popping out if you restart a scenario a lot - for example if on your 1st level up you get proposed intel and Tactics, on your 2nd you're always gonna get Scouting - stuff like that). Art Rolling though to my knowledge is unrelated to that process, so manipulating that would probably be much more elaborate in nature.

Talking about "prep phase" solution isn't going to work, until we can find such preparations, if that ever happens at all.

There is no easy way to fork on windows, without rewriting whole programm. But maybe some linux users may use they magic to check game logic on the moment of creating level and bake it into some docker/another container for shared use.