Not so much a question as much as much as brainstorming so new players may see this in case they needed to figure out how to deal with the topic.

I've only played the Keepers campaign in its entirety so far, but these are the maps I've found work best for such. The key here is that they need to be lengthy maps that may more or less allow you to take your time, explore and slowly but surely conquer your enemies. Note that getting enough mana income is crucial. That involves taking Spellcasting perks with your character as well as additional heroes, and otherwise claiming both plain magic nodes and nodes linked to your schools of magic.

Halfling Arc: United Cities. This is a VERY extensive map which will give you time to get your bearings. In the meantime, researching the spells you have until you got them all is not a bad idea. Recruiting Devon Strongberry in the previous "Northern Trade Route" mission and grabbing Spellcasting perks with him is likely a good idea, but try to use him for combat as little as possible so he doesn't level up too fast and you have trouble carrying him over to later missions. Also helps if you have 2 or 3 spheres in Air magic, because of the Air Node not too far from your location.

Dwarf Arc: Earthquake. Equally as extensive as "United Cities," If not more.

Lizard Arc: Practically any of them except the Southern Marsh. As before, very extensive.

Elf Arc: Isle of Last Goodbyes. This isn't so much extensive as much as beating back the Undead fast enough can let you stall them from progressing and then just take your sweet merry time. Having 3 or 4 spheres in Life is helpful if you claim the Life Node. I remember there also being a Water Node, but I can't remember if there was also an Air Node in the east/southeast portion of the map or if that was a later mission in his arc.

Highmen Arc: Unicorn Valley or The Pass of Grief. Both very extensive.

Undead Arc: Any except Opposition. The only reason I exclude that one is that you have to be on your toes for that one.

If all else fails, you can utilize the final campaign, but it's only recommended if you're with the Highmen since they're allied with the Humans and you don't have to worry about being harassed by their Air Galleys, which the AI tends to make VERY AGGRESSIVE and expansionistic.

Anybody else?

I usually research as many spells as I can on the first map of either campaign.

Dominate the map and lock the opposing leader in one city. Sit around researching until you have everything from the three starting spheres (including cosmic). This also allows you to get the maximum level (3? I think) for your leader.

Anyone else have any input? Feel free to comment if it will help other players. :)

I think everyone does it the way Thereunto described. Whenever you play a new race you also get new spheres and therefore new spell research options, so the first scenario you play as a new race you put the power distribution at 100% mana, win the scenario but leave one leader alive, then research all new spells, and only then kill the last leader. This way you effectively have all mana available for spellcasting at all times.

This thread makes me feel vindicated in my opinion that research costs are too high. Good thing it's easy to edit them!
Plus, I think th next map I make might be another one where leaders don't have any spheres, can only research Cosmos spells, and must conquer certain 'Vaults' to learn spells from the wizard towers there.

I don't quite understand how you came to that conclusion. The spell research exploit exists in the campaign regardless of how expensive spells are to research, and it exists exclusively in the campaign. Multiplayer maps have nothing to do with this whatsoever.

By the way, I'm fairly certain you can't set up a map with fewer than 3 spheres.

It sounds like people feel they need to stick around on a map to research spells, instead of just playing normally. That's how it felt to me too.
As for spheres, I think leaders have to formally have 3, 5, or 7 sphere picks? but you can just not assign them, and play without customised leaders.