Replica exchange molecular dynamics simulations are one of the most popular approaches to enhance conformational sampling of molecular systems. Applications range from protein folding to protein-protein or other host-guest interactions, as well as binding free energy calculations. While these methods are computationally expensive, highly accurate results can be obtained. We recently developed TIGER2hs, an improved version of the temperature intervals with global exchange of replicas (TIGER2) algorithm. This method combines the replica-based enhanced sampling in an explicit solvent with a hybrid solvent energy evaluation. During the exchange attempts, bulk water is replaced by an implicit solvent model, allowing sampling with significantly less replicas than parallel tempering (REMD). This enables accurate enhanced sampling calculations with only a fraction of computational resources compared to REMD. Our latest results highlight several issues with sampling imbalance and parameter sensitivity within the original TIGER2 exchange algorithms that affect the overall state populations. A high sensitivity on replica number and maximum temperature is eliminated by changing to a pairwise exchange kernel (PE) without additional sorting. Simulations are controlled by adjusting the average temperature change per exchange ⟨ΔT/χ⟩ to below 30 K to mimic a controlled temperature mixing of replicas similar to REMD. Thus, this parameter provides an applicable property for selecting combinations of replica number and maximum temperature to adjust simulations for best accuracy, with flexible resource investment. This increases the robustness of the method and ensures results in excellent agreement with REMD, as demonstrated for three different peptides.