Hi,
maybe with stereo the camera would be able to track farther features (with Vis/EstimationType=1). However, such environments are very challenging for visual odometry approaches, as they would give either poor pose estimation or just cannot compute one. If you have a robot, you may use or combine with wheel odometry. For example in the environment of this
tutorial, the Map 5 has long featureless corridors like that, but the robot is using only wheel odometry (with optionally long-range LiDAR for ICP refining).
cheers,
Mathieu