Fix: Bump gripper and GetImage timeouts to stabilize CI

[JWhitleyWork]

Summary

• Bumps MoveGripperAction timeout from 15s → 30s in close_gripper.xml and open_gripper.xml.
• Bumps GetImage message_timeout_sec from 5s → 15s in the shared picknik_ur_base_config Segment Image from No Negative Text Prompt Subtree.

Why

CI on main started flaking ~24h after the MuJoCo 3.2.7 → 3.6.0 upgrade in moveit_pro (6eedef88a5, Apr 14). The new constraint solver is heavier per step, so on CI hardware the sim runs slower than realtime and emits Mujoco model timestep not running in realtime. Increase the model timestep. warnings. Two previously-borderline timeouts in lab_sim integration tests then fall over:

Failing test	~ frequency	Underlying error
push_button_with_a_trajectory.xml	~9/10	MoveGripperAction Error: ... gripper failed to reach the target position within 15.0s. Current values: position=0.7929
ml_segment_point_cloud.xml	~4/10	GetImage Error: Failed to get next message on topic '/wrist_camera/color': Timed out after 5 seconds

CI history correlates tightly with the upgrade:

Period	Pass	Fail
Apr 8–14 (pre-upgrade)	24	0
Apr 15 (day after upgrade)	4	1
Apr 17	1	7
Apr 20–30	~3	~37

This is a mitigation, not a root-cause fix — moveit_pro#18534 and PR #610's description ("the real test harness infrastructure needs some rethinking ... for physics this tight") call out that the durable answer is rethinking the test harness or coarsening the MuJoCo timestep (currently the default 0.002s / 500Hz; no explicit timestep is set in lab_sim/description/scene.xml). Doing that lives in a follow-up.

Test plan

• Trigger CI workflow on this branch and confirm integration-test-in-studio-container (humble) passes.
• Re-run a few times to verify the historical ~92% failure rate drops.
• Manual smoke test: Push Button With a Trajectory, ML Segment Point Cloud, and Close/Open Gripper Objectives still execute correctly on a dev machine.

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR mitigates recent CI flakiness in lab_sim integration tests by increasing timeouts for gripper actuation and image acquisition to better tolerate slower-than-realtime MuJoCo simulation on CI hardware.

Changes:

• Increased MoveGripperAction timeout from 15s to 30s in lab_sim open/close gripper objectives.
• Increased GetImage message_timeout_sec from 5s to 15s in the shared UR base config segmentation subtree.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated no comments.

File	Description
src/moveit_pro_ur_configs/picknik_ur_base_config/objectives/segment_image_from_no_negative_text_prompt_subtree.xml	Increases GetImage message timeout to reduce image topic timeout flakes.
src/lab_sim/objectives/open_gripper.xml	Increases gripper open timeout to reduce actuation timeout flakes under slow sim.
src/lab_sim/objectives/close_gripper.xml	Increases gripper close timeout to reduce actuation timeout flakes under slow sim.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[nbbrooks]

I agree something prob changed with the Mujoco upgrade, but I doubt increasing the timeout will help that much, or at least, is a bandaide.

I would follow the advice in https://docs.picknik.ai/troubleshooting/Planning%20and%20Control%20Troubleshooting/#movegripperaction-fails-to-reach-target-position to address the core of the issue.

[nbbrooks]

Having to bump GetImage past 5sec also seems pretty wild to me. If the sim is lagging so bad we can't get an image reliably after 5sec I don't think we can trust any of the results (who knows how cpu starved the physics is), so personally I would just make this CI job optional and mostly treat it as noise until this is addressed.

[JWhitleyWork]

Superseded by an Option-A approach: instead of bumping per-objective timeouts, we'll add an opt-in input to the moveit_pro_ci/workspace_integration_test.yaml reusable workflow that patches the MuJoCo <option timestep> for CI runs only. That keeps the slower-than-realtime warning visible on dev hardware (where it's diagnostic) and removes the need for tolerance band-aids here. New PR coming.