How do I make MERGE use deletion vectors?

[orthoxerox]

Delta 3.2.1, Spark 3.5.3 bare metal cluster, data in S3

My 1TB table has deletion vectors enabled:

+---------------------------------------------+
|tableFeatures                                |
+---------------------------------------------+
|[clustering, deletionVectors, domainMetadata]|
+---------------------------------------------+

When I perform a MERGE operation on it (when matched update a single field, when not matched insert all), I get the same job in Spark as when I run the same MERGE without deletion vectors: Delta: Writing modified data - MERGE operation - Rewriting 967 files and fail at about the same task when my Spark cluster runs out of temp space.

The docs say deletion vectors are not used when "a DML command with deletion vectors disabled (by a command flag or a table property) is run on the table", but how can I tell if my command has deletion vectors disabled? And how do I enable it on a command?

UPD: I tested the table with DELETE, and it generated deletion vectors. Then I tried an UPDATE, and got Delta: Rewriting 77 files for UPDATE operation instead of new rows + deletion vectors.

[orthoxerox]

I checked the result of the UPDATE and it generated deletion vectors despite saying "rewriting". The same must apply to MERGE