This change is 

@andy-kimball can you take a look at the change to trading and see if we want to add that index explicitly?

@nvanbenschoten do you have any info on how TPCE runs on master if you try to create the schema from scratch (and thus don't get the auto FK indexes)? I see quite a few plan changes in our test file, but this is only a test with no table stats so it may not be too relevant.

pkg/sql/opt/xform/testdata/external/trading, line 1117 at r1 (raw file):

 │         │    │    │    ├── key: (8-10)
 │         │    │    │    ├── fd: (8-10)-->(11)
 │         │    │    │    ├── scan inventorydetails

This is a substantially worse plan for this query, but I think that's OK. It looks like we'd need to implement #51576 in order to get a better plan without adding an index.

bors r+

Build succeeded:

• GitHub CI (Cockroach)

do you have any info on how TPCE runs on master if you try to create the schema from scratch (and thus don't get the auto FK indexes)? I see quite a few plan changes in our test file, but this is only a test with no table stats so it may not be too relevant.

I don't right now. There have been a good number of changes to TPC-E's schema and queries since they were pulled into these tests, so we'll want to update them once things settle down. In the meantime, I'll be on the lookout for auto fk indexes that were actually important next time I spin up the workload.