fix(rtc): validate trained prefix capacity

docs/source/inference.mdx

@@ -280,7 +280,7 @@ lerobot-rollout \
 | `--inference.rtc.mode`                      | `guided` (default) or trained-prefix `trained` for compatible Pi052 checkpoints |
 | `--inference.rtc.max_guidance_weight`       | Consistency enforcement strength (default: varies by policy)                    |
 | `--inference.rtc.prefix_attention_schedule` | Blend schedule: `LINEAR`, `EXP`, `ONES`, `ZEROS`                                |
-| `--inference.queue_threshold`               | Max queue size before backpressure (default: 30)                                |
+| `--inference.queue_threshold`               | Backpressure threshold; trained RTC requires at least its maximum delay         |
 
 See the [Real-Time Chunking](./rtc) guide for details on tuning RTC parameters.
 

docs/source/pi052.mdx

@@ -152,8 +152,9 @@ than `chunk_size`. Choose it to cover the largest inference latency expected at
 deployment: at 50 Hz, for example, 10 steps correspond to 200 ms. A delay of
 zero is included in the uniform sampling distribution, so the checkpoint also
 continues to receive ordinary flow-matching examples. Set rollout's
-`inference.rtc.execution_horizon` to at least this maximum so the previous
-chunk cache retains enough actions to construct every supported prefix.
+`inference.rtc.execution_horizon` and `inference.queue_threshold` to at least
+this maximum so inference starts early enough and the previous chunk retains
+every action needed for the committed prefix.
 
 Run the resulting checkpoint with the asynchronous `lerobot-rollout` backend
 and select the trained-prefix path explicitly:

docs/source/rtc.mdx

@@ -97,6 +97,10 @@ for step in range(num_steps):
 - `guided` (default) applies the original Jacobian guidance during denoising and works with ordinary flow-matching checkpoints.
 - `trained` hard-inpaints the previous chunk's prefix with per-action flow timesteps. It currently requires a Pi052 checkpoint trained with `policy.rtc_training_max_delay > 0` and avoids the guidance backward pass.
 
+For trained mode, both `execution_horizon` and the rollout backend's
+`inference.queue_threshold` must be at least the checkpoint's
+`rtc_training_max_delay`; rollout validates this before connecting the robot.
+
 **`execution_horizon`**: How many timesteps from the previous chunk to maintain consistency with. Higher values mean smoother transitions but potentially less reactivity.
 
 Typical values: 8-12 steps
@@ -129,6 +133,10 @@ python examples/rtc/eval_dataset.py \
     --device=cuda
 ```
 
+Add `--rtc.mode=trained` when evaluating a compatible training-time RTC Pi052
+checkpoint. Unsupported policies reject trained mode instead of falling back to
+guided RTC.
+
 The script generates a visualization of the denoising process, comparing standard generation (left) with RTC (right). In the RTC plots, you can see how the first few steps (blue/purple lines) are guided to match the red ground truth trajectory (previous chunk's tail), ensuring a smooth transition between chunks.
 
 <p align="center">