11.48N+1 at bulk scale is catastrophic

With normal API work:

100 records
× 1 extra query

is annoying.

With bulk:

1,000,000 records
× 1 query each

is disastrous.

So before large jobs:

pre-fetch lookup data
batch queries
create in-memory/file-store mapping where reasonable
use external IDs

rather than query one-by-one.