When using a LOP HDA in the HDA processor, the output (using "write output in the HDAprocessor) is being flattened. Though this is not always desired. In my case, the HDA consists of multiple point instancers, using all kinds of vegetation assets. These are referenced, but as the output is flattened, the prototypes are baked into the output, duplicating a lot of data.

A workaround I've tried is to use an USD rop inside the HDA which gets triggered by the execute parameter, but in that case the HDA processor doesn't track the output file and will always re-cook, instead of caching from disk when the output is already available.

So two questions: Does anyone have a different solution to write out an unflattened stage from the HDA processor?
And: Can I manually define expected outputs for the HDA processor, just like the ROP fetch supports with "Output Paths from: Custom file list"?

Sent the same questions to support and got a suggestion for a workaround:

"The HDA Processor currently doesn't have any options for controlling how .usd files are are written out. We can turn this issue into an RFE for that, but the best workaround now would be to create an instance of your HDA in your scene and use a ROP USD Output TOP to write out USD files instead. That TOP node has the same options for writing out .usd files as the equivalent ROP -- for example flattening, output processors, etc."

An RFE (#149757) is logged to add USD save options to the HDA processor.

joostkonemann
Sent the same questions to support and got a suggestion for a workaround:

"The HDA Processor currently doesn't have any options for controlling how .usd files are are written out. We can turn this issue into an RFE for that, but the best workaround now would be to create an instance of your HDA in your scene and use a ROP USD Output TOP to write out USD files instead. That TOP node has the same options for writing out .usd files as the equivalent ROP -- for example flattening, output processors, etc."

An RFE (#149757) is logged to add USD save options to the HDA processor.

We have a publisher in place that handles a significant amount of work on top of the incoming USDs. We have a stage inside the TOP HDA, where we cook the output layer as we prefer, typically the same way you do, without flattening. In other words, if you really need your HDA, you can embed it in another one that will do all the work.