Localization and mapping stand as fundamental capabilities for unmanned systems to accomplish intricate tasks. In the domain of multiple robots, challenges persist in Collaborative Simultaneous Localization and Mapping (C-SLAM). Particularly, when robots revisit locations previously traversed by others, issues arise in loop closure detection and establishing inter-robot connections due to lateral lane variations. To tackle this challenge, we introduce DiCar-SLAM, an optimized distributed multi-robot SLAM system using 3D lidar and inertial sensors. DiCar-SLAM leverages a novel 3D point cloud feature descriptor named Enhanced Scan Context, tailored for distributed C-SLAM. This descriptor ensures rotational invariance while fortifying lateral robustness, thereby facilitating loop closure detection across multiple robots navigating different lanes within the same scene. We extensively evaluate and compare DiCar-SLAM with other C-SLAM baselines above several collaborative SLAM datasets. Our quantitative assessments demonstrate the superior suitability of DiCar-SLAM for collaborative localization and mapping scenarios involving lateral lane changes.