Intelligent Reconfigurable Surfaces (IRS) are crucial for overcoming challenges in coverage, capacity, and energy efficiency beyond 5G (B5G). The classical IRS architecture, employing a diagonal phase shift matrix, hampers effective passive beamforming manipulation. To unlock its full potential, Beyond Diagonal IRS (BD-IRS or IRS 2.0) emerges as a revolutionary member, transcending limitations of the diagonal IRS. This paper introduces BD-IRS deployed on unmanned aerial vehicles (BD-IRS-UAV) in Mobile Edge Computing (MEC) networks. Here, users offload tasks to the MEC server due to limited resources and finite battery life. The objective is to minimize worst-case system latency by optimizing BD-IRS-UAV deployment, local and edge computational resource allocation, task segmentation, power allocation, and received beamforming vector. The resulting non-convex/non-linear NP-hard optimization problem is intricate, prompting division into two subproblems: 1) BD-IRS-UAV deployment, local and edge computational resources, and task segmentation, and 2) power allocation, received beamforming, and phase shift design. Standard optimization methods efficiently solve each subproblem. Monte Carlo simulations provide numerical results, comparing the proposed BD-IRS-UAV-enabled MEC optimization framework with various benchmarks. Performance evaluations include comparisons with fully-connected and group-connected architectures, single-connected diagonal IRS, and binary offloading, edge computation, fixed computation, and local computation frameworks. Results show a 7.25% lower latency and a 17.77% improvement in data rate with BD-IRS compared to conventional diagonal IRS systems, demonstrating the effectiveness of the proposed optimization framework.