mana screw and flood in their extremes can be bad because they reduce gameplay options to the point where you can't do anything (although having somewhat less or somewhat more mana than is optimal are both good variance and more common outcomes). its good to have to think about how to play from behind, but that only applies if you have any options at all.
Yeah.
The strictly optimal amount of lands to have is 1 in the opener, 1 every turn for the next X turns where X is anywhere from 1 to 9 depending on the deck, and then none after. If you have more, you get less choices for spells and sometimes no spells at all; if you have less, you skip land drops and sometimes are gated from higher-costing cards.
The interesting thing here is that having fewer choices is often barely noticible but having no spells is devastating; and that missing land drops is on it's own often barely noticible (not always, more noticible in faster formats) but being gated from higher-cost cards is devastating, especially if the cost is 3 or 4. When it comes to flood and screw variance, the middle ground exists, but it's not very noticible, and I would say the impace of middle ground variance from the number of lands is less than the impact of middle ground variance from costs and effects of spells drawn.
HS Basic/Classic, arguably the peak of Hearthstone, didn't use that many random effects and there was still quite enough variance just from CMCs and having/not having the right cards and whatnot. Hearthstone's system has plenty of variance, and I'd say a bigger problem for it is that it still keeps Magic's card determinism (if you drew a card, your only options are playing it or ignoring it until you play it) and it introduced lots of mechanics trying to address the resulting lack of choices - Discover and Tradeable are now both evergreen, for one.
"You can play red cards as Mountains and blue cards as Islands" is so obviously the best choice for the mana system - it removes mana screw and flood while keeping slightly tamer color screw, it eliminates card determinsm in a way that turns every single card into a decision (but not an overbearingly complex one! it's binary and one of the halves is super consistent) and eliminates the need for the ugly hack of sideboarding, it keeps open the possibilities of rewarding people for comitting to monocolor or almost monocolor with multi-pip cards or wantonly mixing the colors; and finally, it opens up cool design space with lands and nonlands organically transforming into each other.
Obviously, when it comes to art and entertainment, "the best choice" is fundamentally subjective, but I'd say that here, it's about as subjective as "touching a hot stove is not a good choice". I honestly don't know why isn't almost every new game using this.