Am I correct in the understanding, that hibernate Session cache/1. level cache only caches entities based on the PK-mapped column. In my case, because I use JPA, the @Id annotated property?

Reason I ask is, that I have mappings between entities on non-PK mapped entities (but they have a DB unique-constraint, so they ARE unique), and it would be über cool, if hibernate could cache on that value also. I use @JoinColumn to tell it which to join on.

The database table design is this:

TABLE parent (
ID INT (PK column),
OTHER_ID VARCHAR (UNQ, but not PK column)
)

TABLE child (
PARENT_OTHER_ID VARCHAR (FK to parent.OTHER_ID)
)

And I would like hibernate to cache "parent" instances based on BOTH "ID" and "OTHER_ID" in the 1. level cache.

While I wrote this up, it kind of sounded like impossible to me, but is there any cool stuff I need to know about that could help me out here? (aside from changing my datamodel, that is).

Yes, that's right.
If you want to have cached elements by @NaturalId or relation collections
then you must consider to use the 2nd level cache.

It was my understanding that the second-level cache would only index on id too. Am I wrong in that?

that's correct, second level cache will also index on @id.

However, I think you can use query cache, if you want to query out a collection using some other column than Id.

You are partially right and at same time wrong:
-if you only enable second-level cache without enabling second-level-query-cache, then the does index only on id.
-if you enable the second-level-query-cache too, then the cache can also index collections and query results.
So in order to cache lookup queries on @NaturalId you would have also enable the second-level-query-cache.