Hello, We’re integrating XGen via XGenRenderAPI for our own renderer. Parsing works, but when artists edit XGen parameters and we re-parse in the same Maya session, XGen emits repeated errors: _“clumping guide ID larger than number in xuv file”_ Deleting **.xuv** before parse avoids the error. In Arnold, when caches are stale, we see: [xgen] Batch mode tests failed! proceeding with Xgeneration... using live scene data — which appears to force a “live scene data” path and avoids stale clump caches. What we’re doing: auto* patch = XGenRenderAPI::PatchRenderer::init(&cb, cmd.c_str()); auto* par = XGenRenderAPI::ParallelRenderer::init(patch); cb implements ProceduralCallbacks (flush/log/get* etc.). Even with an empty flush(), the stale .xuv issue persists unless we delete them first. The question is: * **How to trigger “live scene data”** reliably from a third-party renderer? Is there a documented **cmd string** or flag for PatchRenderer::init(...) to bypass batch caches and use live scene data (like MtoA’s fallback), or is the intended approach to fail batch validation and let XGen switch internally? * **Best-practice cache handling** : What is Autodesk’s recommended approach to avoid clump ID desync without deleting user .xuv maps? For example, does XGen provide a supported API call to **regenerate clumping maps** programmatically or to **redirect** the clump “Points Dir” per-parse to an isolated location? Thank you very much.