Load-stress test is the experimental performance analysis in extreme traffic and density conditions, routinely required to validate any innovative radio access solution. This paper focuses on load-stress test specifically designed for the two-hop architecture that enables the onboard connectivity in high speed trains (HSTs). The load-stress condition of train-to-infrastructure communication for a massive number of on-board users (UEs) is very challenging, as it needs to account for extreme conditions and complex testing environment. Load-stress method proposed in this paper is for ground-network supporting onboard wireless connectivity in HST, it is validated for commercial evolved NodeBs (eNBs) from Long Term Evolution (LTE) cellular networks (Rel-11). The in-Lab experimental setup is arranged by virtualizing multiple eNBs serving multiple cells, arranged sequentially along a line to simulate the HST track with a massive number of active onboard UEs. The focus of the experimental load-stress test is the analysis of the impact of Doppler and frequent interruptions caused by the frequent handovers (HOs) of multiple consecutive groups of UEs deployed in HST carriages at the speed of 300km/h. The HO interruption time is characterized statistically based on the number of active UEe. The consequent impairments on the experienced QoS for high-throughput and low-latency services such as FTP and VoLTE are verified. This paper validates experimentally the traffic and HO latency improvements (approx. x3) in multi-cell access scheme, when the coverage of every single carriage is augmented by fixed directional antennas to offload the UEs toward far-away eNBs along the train track.