A hunt answers to the id the game publishes — ab-bonelord-hunt — and to its slug, and
a name is slugged before it is tried, so /hunts/braindeath-caves and
/hunts/Braindeath Caves are the same hunt. The id wins where the two collide.

Both handles earn their place on every row of this track: the game's id is the place
with a -hunt suffix where the name is the place alone, so the two are different
strings on all 60 hunts.

Its loot table comes with it unasked, where the listing takes ?include=loot.item:
heaviest chance first and the item id breaking a tie, and a line may carry an
item_client_id with a null item, which is the shape of a database whose item track
has not run. A hunt never lists an item twice — measured, 1,223 distinct hunt-and-item
pairs over 1,223 lines, where 18 of the 113 creatures repeat one in their own loot — so
a client may key this table on the item id where a creature's loot must not be keyed.

Three field names moved in this release, which is a breaking change. monsters
is creatures, tiers[].monster_count is tiers[].creature_count, and
tiers[].item_client_ids is tiers[].loot_item_client_ids. Same values, same order,
same nesting; the tables keep the word monster and no filter or sort took it as a
name. creature_count is still not the length of creature_keys beside it.

A tier's name is now a difficulty object, which is a breaking change.
tiers[].name was the bare label and is tiers[].difficulty.name now, with
difficulty.slug beside it and difficulty.rank — 1 to 5 — saying where the rung sits
on the game's whole ladder. position is unchanged and answers a different question:
it counts from 0 within this hunt, so a hunt that starts at Reckless has a
position 0 whose rank is 3.