XMP Toolkit SDK -- DOM API?

I notice that the XMP Toolkit SDK has a relatively new DOM API, as documented in XMPAddendumProgrammersGuide.pdf. This lets you navigate a node-based object model e.g.

auto registry = IDOMImplementationRegistry::GetDOMImplementationRegistry();
auto serializer = registry->GetSerializer("rdf");

auto metadata = IMetadata::CreateMetadata();
auto simpleNode = ISimpleNode::CreateSimpleNode(kXMP_NS_XMP, AdobeXMPCommon::npos, "CreatorTool", AdobeXMPCommon::npos);
if (simpleNode) {
    simpleNode->SetValue("Test", AdobeXMPCommon::npos);
    metadata->InsertNode(simpleNode);
}

auto defaultMap = INameSpacePrefixMap::GetDefaultNameSpacePrefixMap();
auto str = serializer->Serialize(metadata, defaultMap);
               
auto xx = str->c_str();
std::cout << xx;

I don't see much docs or use online, so I'd to know more about it:

• Does anyone here use it in preference to the older XMPCore interfaces? What was your experience like?
• Why would I choose the DOM API over the older API? It seems the older API didn't have a distinct object model, so navigation there has to go through paths and enumerating tags is an all-at-once affair.
• Does the DOM API faster or use less memory over the older API?
• Does the DOM API do lazy or incremental parsing or is the entire object graph set up during the initial parse?
• Can the Adobe engineers/managers who wrote the DOM API share some of the design considerations and motivations for writing this? Will this be developed further?

Thanks, folks.