The research project the data originated from: MoS2 nanoswitches have shown superb ultra-low switching energies without excessive leakage currents. However, the debate about the origin and volatility of electrical switching is unresolved due to the lack of adequate nano-imaging of devices in operando. Here, we combine three optical techniques to perform the first non-invasive in situ characterization of nanosized MoS2 devices. Our study reveals volatile threshold resistive switching due to the intercalation of metallic atoms from electrodes directly between Mo and S atoms, without the assistance of sulfur vacancies. We observe a ‘semi-memristive’ effect driven by an organic adlayer adjacent to MoS2, which suggests that non-volatility can be achieved by careful interface engineering. Our findings provide a crucial understanding of nano-process in vertically biased MoS2 nanosheets, which opens new routes to conscious engineering and optimization of 2D electronics. $$ \ $$ The dataset usage instructions: Names of data files correspond to the names of Figures whose data they contain. Files: Figure_1f.002 and Figure_S3a.013_2 are rough AFM scans used for Figures 1f and S3a, respectively. To process the data, please download the Gwyddion software and follow the instructions from: http://gwyddion.net/. Optical and electrical data for Figures 1b-c, 2-3, 4a, 4d, 5a, S2a-b, S5 can be processed using any software for data analysis, such as Spyder (https://www.spyder-ide.org/). All optical data was pre-processed with background subtraction. Additionally, dark field scattering data (Fig. 1b, 4a, 4d, S2b) were reference normalized. Lorentzian curves were fitted to data (Fig. 1c, 2c, 3c) using Spyder software. Finally, Lorentzian fits of the photoluminescence response were subtracted to achieve clear Raman spectra (Fig. 1c, 3c insert). For all electrical data current was recalculated to current density by dividing measured current by electrode’s area (706.5 nm2). The outcome of FDTD simulations was exported to .txt format and provided in files for Figures 4b-c, S2c, and S4. It can be accessed by any software for data analysis, such as Spyder (https://www.spyder-ide.org/). $$ \ $$ The data collection methods: The optical and electrical data (files for Figures 1b-c, 2-3, 4a, 4d, 5a, S2a-b, S5) were collected using self-made setup as described in: https://doi.org/10.1002/adma.202209968, 5 Experimental Section -> Electrical Setup, Optical Setup. The setup was operated with self-written software in Python 3. FDTD simulations (files for Figures 4b-c, S2c, and S4) were performed as described in https://doi.org/10.1002/adma.202209968, 5 Experimental Section -> Electrical Setup, FDTD Simulations. Atomic force microscopy (AFM) scans (files for Figures 1f and S3) were collected with the AFM Multimode Nanoscope III. All data were collected at the University of Cambridge, Department of Materials Science & Metallurgy.