In the pursuit of observing fainter astronomical sources and phenomena, a significant challenge in detector development lies in ensuring that these devices can detect each individual photon they receive. By amplifying each incoming photon by several orders of magnitude, electron-multiplying charge-coupled devices (EMCCDs) offer a promising solution to meet this challenge. Although these detectors boast impressive potential, they can be intricate, requiring precise optimization and fine-tuning of their parameters to unlock their full capabilities in the photon-starved regime. The Faint Intergalactic-medium Redshifted Emission Balloon (FIREBall-2) is a stratospheric project that aims to detect and map the low surface brightness environment of galaxies in the ultraviolet (UV) at z∼0.7. As a technology demonstrator for photon-starved astronomy and to advance the technology readiness level of UV EMCCDs, the instrument uses a Teledyne-e2v (T-e2v) EMCCD delta-doped by the Jet Propulsion Laboratory, combined with a NüVü controller. To analyze the detector data and retrieve the device noise contributions, we developed a comprehensive EMCCD model along with DS9 analysis tools to compare the model with the actual data under diverse operating conditions. This allowed us to examine the current performance and limitations of these devices both on the ground and in the stratospheric environment, to unravel the intricacies of these detectors. In addition, we will discuss the development and implementation of an exposure time calculator designed to optimize the end-to-end signal-to-noise ratio under diverse conditions and analyze the different trade-offs associated with such devices. This will be used to explore some EMCCD-related issues encountered on FIREBall-2 and present some recent and potential future upgrade strategies (controller upgrade, red-blocking filter, over-spill register implementation, etc.) to mitigate them.