I had similar issues with HLODs. They were sometimes failing to build, and after the build, they were causing validation errors because of invalid GUIDs. I came up with a workflow like this to avoid all the issues:
Delete HLODs.
Submit a CL with deleted HLODs. - without this step, new HLODs were failing to build.
Build HLODs.
Submit a CL with new HLODs.
Restart the editor! - without this step, asset validation was failing because of invalid GUIDs related to HLOD objects.
Launch the session, memory calc, or whatever you need.
It is very time-consuming and strenuous, but in UEFN you have to get used to that
i am by no means an expert so if im wrong in any part of this please correct me, but i have had to deal with hlods and heres what ive learned…
if your hardware isnt the cause one thing to look at is if your hierarchy actually follows the “rules” for hlods, specifically the “hierarchical” part. if youre copy/pasting things all over as single objects and/or unparenting them from prefabs while still flagged as hlod youre going to run into these memory issues every time, especially on big maps where youve done this a lot. you want all your objects that make up a single structure/building to be under the same root object otherwise youre generating unique hlods for each individual piece, which defeats the purpose and can cause these memory errors. a good way to think of it is to think of a city… all walls are children to the building, all buildings are children to the city block, all city blocks are children to the city itself
another thing to consider is what is actually being flagged for hlod. small objects inside structures that cant be seen from outside the structure dont need hlod just use streaming. hlods are most beneficial for large structures visible from a distance. sometimes theres smaller objects that you can get away with just streaming. single large objects (like background mountains) dont need hlod or streaming so you could untick both
hlods are powerful, but they are also tricky/time consuming to master