I’ve attached a repo video above as well as screenshots to show how to see it even without looking into the hmd in the spectator preview
You do know that Steamlink is simply a streaming app directly from the Quest that uses SteamVR as the OpenXR runtime right??? Its no different than testing SteamVR as the OpenXR runtime via a vive pro, valve index or quest link cable running with Steamvr as the default runtime…
In regards to your question… I had already went back and updated that I also tested with native Meta OpenXR and SteamVR native without SteamLink. I’m not sure what you mean by you can’t reproduce it? Can’t reproduce what? That 5.5 preview puts raytracing and skincache into the forward rendering VRTemplate? Or that stereo instancing is broken? I left very detailed steps on the order of operations I followed but it’s very simple to repo. Open the VRTemplate and set all lights to movable and clear the lightmap data by rebuilding or deleting the umap build data file then test in “any runtime” Steamlink, openxr, meta quest link… Or just hit play without changing anything and close your left eye and then your right eye while watching the shadows cast from your hands. This isn’t a runtime issue its a shader .ush issue with how the shadow projection is handled which I have patched here for every hotfix since 5.4.0 released… additionally people have been using if they want to move past 5.3 (with forward rendering) and have replied here with thank you. You can also see that I’ve been in contact with the original commit owner and also provided a link to the pointlight fix here as a discussion on the original CL committed to UE5 main. As far as CSM vs VSM deferred and forward I outline everything very detailed above so that there isn’t confusion. There are 2 different rendering paths, 2 different shadow methods, and considerations for each have to take into account instanced stereo and VSM and CSM (if your in deferred). The above video outlines the simplest repo which is literally a default template with zero changes other than adding the spectator view enum to show both eyes. I took the time to show shadow casting with default template settings (Broken) and then with instanced stereo disabled (works).
In the most respectful way possible I simply ask after all of these votes and after all the work I’ve done to document commit to the engine, create a git repo with code changes, record several videos… Why is the burden of proof still on us after a 2 line response that “We can’t reproduce it” Can you clarify if you can’t reproduce it in UE5.5 main on source or 5.5 preview (launcher)? Can I kindly ask that you take the time to show a video of it actually working? I find that only fair if your claiming it is something 30+ of us on this forum and 100+ on other forums are doing wrong? Especially when I’ve written code to correct the issue and you can simply see numerous post stating this isn’t an issue in 5.3. This is a massive convo in VRE discord, Oculus Start, and Unreal Engine Source which I myself have pinged you in there on each update to this forum. Please accept my apologies if this feels like its directed personally toward you as its not. I know your extremely busy. I met you in new orleans and we geeked out for like an hour back when UE5 released. I’m simply trying to get visibility on this issue and help our fellow developers and patching 4 hotfixes with 5.5 coming and still seeming to have this issue is a tad exhausting. It’s a plea for help from the inside at this point 
As far as repoducing the ini issue you don’t even have to open the project… just make a VRTemplate from UE5.4 and another one from 5.5 and go look at the difference between the two defaultengine.ini render settings.
Also quoted in this forum and riddle through many others are plenty of reproduction reports.
