Clarify generic ZeroGPU runtime failures

app.py

@@ -559,8 +559,8 @@ def ui_inverse_design(n_generate: float, mode: str, load_idx: float, *values: fl
         generated = _run_generation_model(cond_coeffs, n_generate_int)
     except Exception as exc:
         exc_text = str(exc)
-        if "No CUDA GPUs are available" in exc_text:
-            title = "ZeroGPU allocation failed before CUDA became available."
+        if "No CUDA GPUs are available" in exc_text or exc_text.strip("'\"") == "RuntimeError":
+            title = "ZeroGPU allocation failed or returned a generic backend error."
         elif "ZeroGPU quota" in exc_text or "quota" in exc_text.lower():
             title = "ZeroGPU quota is exhausted or unavailable."
         else:
@@ -568,9 +568,10 @@ def ui_inverse_design(n_generate: float, mode: str, load_idx: float, *values: fl
         message = (
             f"{title}\n\n"
             f"Error: {type(exc).__name__}: {exc_text}\n\n"
-            "If this is running on ZeroGPU, wait and retry if the GPU queue is busy. "
-            "If the account or organization has exhausted ZeroGPU quota/credits, switch the Space to "
-            "`Nvidia T4 small` or restore quota before retrying."
+            "Hugging Face's native ZeroGPU waiting popup only appears while a GPU allocation is actively queued. "
+            "If the ZeroGPU backend immediately fails, times out, or returns a generic `RuntimeError`, the app may "
+            "only receive this exception after waiting. Retry after quota is restored, authenticate for more quota, "
+            "or switch the Space hardware to `Nvidia T4 small`."
         )
         return (
             req_plot,