Existing studies on progressive collapse of reinforced concrete (RC) flat plate structures have mainly focused on single column loss scenarios. However, accidental events, such as earthquakes, blasts or vehicle collisions, may cause a more severe initial damage beyond the loss of only one column. To fill this gap in knowledge, this study reports two quasi-static large-displacement experimental tests on two nominally identical 1/3-scale, 2 × 2-bay RC flat plate substructures under an edge-column (S-E) and an edge-interior-column (S-EI) removal scenarios. Two types of uniformly distributed loads (UDLs) were applied to the slab in the two tests: (1) an increased UDL during the tests on the bays adjacent to the removed column(s) and (2) a constant 5 kPa UDL elsewhere to simulate the design live load. In both tests, punching shear failures were observed and led to the first peak loads (FPLs) which were all followed by a sharp load drop. Subsequently, the load increased again to reach the post-failure peak loads (PPLs). The experimental results showed that the PPL was 9.6% higher than the FPL in S-E and 81.9% higher in S-EI. Relative to S-E at the PPL, S-EI presented a more ductile structural behaviour, in which the applied load at the PPL was 26.9% lower, but with a 112% larger displacement, than S-E. In both tests, tensile membrane actions were observed at large deformations and found to be essential in developing post-failure capacities. Such resisting mechanism cannot be neglected when investigating progressive collapse of RC flat plate structures. The yield line theory was found to overestimate the flexural capacity of the two tests.