Abstract <p>Hybrid molecular dynamics/grand canonical Monte Carlo (MD/GCMC) simulations are increasingly employed to study adsorption-induced structural transitions in flexible metal–organic frameworks (MOFs). While these approaches can capture guest-triggered transformations such as breathing phenomenon or negative gas adsorption, their fundamental validity hinges on ergodicity—a requirement rarely scrutinized. Here, I critically examine the ergodic properties of MD/GCMC and related hybrid schemes. I show that although such methods preserve statistical ergodicity by design, they break dynamical ergodicity due to discontinuities introduced by particle insertion and deletion moves. This breakdown has consequences, including dependence on initial conditions or biased sampling, and loss of physical interpretability of time correlations. I compare MD/GCMC, osmotic molecular dynamics, and hybrid osmotic Monte Carlo, and identify the conditions under which ergodicity is breakdown. This analysis establishes criteria for the reliable use of hybrid simulations and highlights the need for physically consistent alternatives that explicitly couple MOFs to real gas reservoirs.</p> Graphical abstract <p></p>