When reading the docs at http://jackrabbit.apache.org/filevault/importmode.html I would assume that for ImportMode.MERGE the merging happens on node level (i.e. indepent of the actual serialization format). Unfortunately this is not the case: If I have a serialized docview .content.xml file covering the current node and three levels below the merging is not done if the entry node level does already exist in the repository, although not all child nodes are in the repository yet.
This is due to the guard in https://github.com/apache/jackrabbit-filevault/blob/8b2fedaf329b3bf4a049e41563c0bc83487406f7/vault-core/src/main/java/org/apache/jackrabbit/vault/fs/impl/io/FileArtifactHandler.java#L226. which prevents the DocViewSAXImporter from kicking in, although there are indeed subnodes in the .content.xml which are not yet in the repository.
Please either fix this behaviour of make the documentation at http://jackrabbit.apache.org/filevault/importmode.html clearer because it explicitly says there:
It is important to note, that the import mode always operates on entire nodes and subtrees...