I agree it's complicated to figure out licensing in the Software Asset Management use case with an on-premise UCMDB deployment.  To be clear, we have drastically simplified this process when you use a SaaS deployment (with "asset discovery" instead of "express discovery" , but the confusion remains with on-premise.

It's true that Express discovery will detect everything you need to know about the VMs themselves in a virtual environment including underlying virtual hardware and the installed software.  This is because Express discovery logs into the server directly and makes it's queries.  Express discovery does not have an understanding of the underlying virtualization hosting environment though - it sees the VM as a standalone server.  That is, while the VM may have 2 virtual CPUs, it has no visibility to how those 2 virtual CPUs translate into physical CPUs on its ESX host.  Additionally, basic discovery of the ESX host will also expose its CPUs, but again, there would be no visibility to how those physical CPUs are mapped to the virtual CPUs on the VMs it hosts.  To get this data requires the full discovery of vCenter or vROps.

But, for purposes of straightforward Software Asset Management, do you need that level of detail?  From what I understand, you only need that level of detail if you are tracking Oracle or MSSQL license management in SAM (which it does appear you are doing).  Also, the Oracle LMS discovery would require this level of detail as well to generate the Oracle measurement script data.

Discovery of details for running software does require a full unit (Premium discovery) per system.  That means that any system you discover where you need details for software like Oracle and MS SQL will require that the "Host Resources and Applications" (HRA) discovery be run against that server (thus consuming 1 unit for discovery of that server).  If you only care about this information for systems where Oracle or MS SQL are installed, you can tweak the discovery trigger query for each of those jobs to add a dependency on the existence of an Oracle or MSSQL installed software CI.  In this way, you would be running the HRA job only on systems where this software is installed and therefore consuming fewer Premium units of discovery. 

Yes, there is a document that outlines in detail all the discovery jobs that are part of Universal Discovery and CMDB.  This document is available as a download in the same place where you download the UCMDB bits.  The title of the package would be something like "OpenText_Universal_Discovery_License_and_Unit_Consumption_Guides-24.4.0.zip".

For security reasons, I can't post this document to a site that is open to the internet (hence the reason the download is protected by a credential as part of SLD).  However, reach out to me directly at bdyck@opentext.com and I can send this document to you.