HydraIssues

Split out of #615 on 2026-09-01. The build pipeline work must not be held up by this.

What happened

The first hydragon build (a copy of UE 5.7's TP_VirtualRealityBP template, cooked on fluffy-dumpling-87 on 2026-09-01) packages and launches, and it renders lit three dimensional geometry. It does NOT render the full template scene.

Evidence: the packaged build was run flat with -nohmd -windowed -ExecCmds="HighResShot 1280x720" through an interactive scheduled task. The capture shows a single lit sphere with a correct light and shadow gradient, on black. The rest of the VR template room is absent.

The log carries, repeatedly:

LogRHI: Error: Depth Stencil Format:
LogRHI: Error: 0x0

What this does and does not prove

Proven: the project cooks, packages, launches, creates a rendering device, loads content, and draws lit geometry with working materials and a directional light. The render path is alive.

Not proven: that the scene is correct. A partial render is not a pass, and this must not be reported as one.

Likely causes, to be worked through

1. The capture happened before the level finished loading and lighting settled. The cheapest thing to rule out first: add a delay or a load-completion gate before HighResShot.
2. The Depth Stencil Format: 0x0 errors point at a real RHI problem, possibly specific to running windowed in a scheduled task session rather than a normal interactive launch.
3. fluffy reports three video controllers, two of them Virtual Display Driver. That is the #428 condition, and it can make an application pick the wrong adapter.
4. Lighting may need a rebuild. VRTemplateMap_BuiltData.uasset ships with the template, but a cook can invalidate it.

Also in scope

Whether the same build renders correctly in the headset over ALVR, in stereo. That is the real acceptance test, and it needs a person wearing quest-head-36afcd. Today that head is offline with a null xr_client_hostname, which cannot be set remotely.

Done when

The packaged hydragon build renders the full scene flat, with no RHI errors in the log, AND a person confirms a correct stereo view in the headset.

Related: #615, #428 (double Virtual Display Driver), #636 (packaging), #626 (XR lanes).