Blood Island is huge, like a couple hundred miles by a couple hundred miles. It's a hexcrawl, in that the parties go from hex to hex and discover what they can. Each hex has either:
1 Nothing. Describe the terrain well.
2 Random monster encounter.
3 Something special (old mine, cave system, ruin, monster plus den to explore, former settlement of their people, former settlement of the enemy, place of mystery or wonder, place with a strange magical effect, special monster, intelligent monster, camp or settlement of an intelligent species/race never before seen to make an ally or enemy of, distinctive natural landmark, magical natural landmark, portal that takes you to somewhere else on the island, portal that takes you to another plane, natural hazard like quicksand or mudpit, a non-natural trap of some sort, entrance to the underdark, etc.)
4 Something special, like above, but which gives clues to unlocking the greater mystery of the island.
I've yet to figure out the odds of each thing on the d100 roll, but they order above is roughly from common to rare.
As a DM, if I want to put a special thing which gives clues somewhere on the island, like say on hex 276, and that clue is somehow important to unlocking the greater mystery, but the party never goes there, and therefore never discovers it, what a waste of time that would be to create the whole complex/dungeon/whatever. It would also annoy the players to have to figure out what hex they haven’t gone to, in order to find the thing that unlocks the riddle so they can conquer the whatever.
At the same time, you don't want to make wherever they go be the next thing you want them to find, based on what you've just created. That's a railroad, in that no matter what the players decide to do, they will end up with the encounter you prepared for that session. That defeats the purpose of the whole sandbox approach to hexmap exploration.
Here's what I've decided to try out: we already randomize monsters, so why not randomize all the other encounters? If I create a half dozen or so of each type of encounter, and number each one, if I roll for what is in the next hex the players encounter and it comes up “abandoned mine,” I go to my collection of half a dozen abandoned mines which I've already mapped out and stocked with monsters and treasure and clues and giant rats and copper pieces and all that cool stuff, and randomly choose one to be the one the players encounter. After that, I never use that encounter again.
If I take this approach, there is no railroad, and no DM time and resources are wasted. Wilderness encounters become a series of random one page encounter areas, basically.
Obviously I reserve the right to re-roll if the encounter doesn’t make sense for either the geographical area, or if it doesn't fit with what the players have already discovered nearby. But I think this approach can work.
It will require a lot of upfront design time on my part, but after each session all I need to do is redesign a few new encounters of whatever encounter types the players explored that session. That way I don't re-use things, the players aren't bored, and it keeps it fresh.
It seems like a good way to make a wilderness hexcrawl interesting for the players, as opposed to a lot of random monster encounter areas, avoids a railroad, and it maximizes DM efficiency and creativity by making sure that whatever I make up is eventually used. The other benefit is that if the party randomly teleports or otherwise finds a way to go to a some random part of the island (never underestimate how good players can fuck up all preconceived notions of what that adventuring session was going to be about), you as DM don't have to have filled in all 6000 hexes in advance to be ready for what they do and where they end up. Also note that this method doesn't prevent a DM from making certain hexes have definite fixed encounters. If I want hex 541 to have the remains of the tower of the arch-mage Felix Unger, it can be there as a permanent static encounter, without interfering with the random hex system above.
What do you think?