Inconsistency between success condition and reward threshold in door_close.

[CEW00]

Hi, I noticed a small inconsistency in SawyerDoorCloseEnvV3 (v2 reward function):

The success condition is (in def evaluate_state):
"success": float(obj_to_target <= 0.08) (line 112)

The reward condition is (in def compute_reward):
if self.reward_function_version == "v2":
_TARGET_RADIUS: float = 0.05 (line 126, 127)

if obj_to_target < _TARGET_RADIUS:
reward = 10 (line 154, 155)

This would mean the environment can give the reward without the success condition being met, which might cause exploitation behaviours.

Am I missing something? Is this an intentional choice?