<p>Many metaheuristic optimisation algorithms applied to the least-cost design of WDNs require extensive parameter tuning and often experience performance degradation as the problem size increases. To overcome these limitations, this study proposes a Hybrid Jaya–Genetic Algorithm (HJaya-GA) that improves the scalability of the Jaya algorithm while requiring only a minimal set of commonly used metaheuristic parameters. The proposed framework incorporates three complementary mechanisms adopted from the literature: (i) a Prescreened Heuristic Sampling Method (PHSM) to direct the initial population toward promising regions of the search space; (ii) Time-Varying Acceleration Coefficients (TVACs) to regulate the transition from exploration to exploitation; and (iii) a GA-inspired adjacency mutation operator to maintain population diversity. Both Jaya and HJaya-GA algorithms are evaluated on three benchmark WDNs of increasing size and complexity using widely adopted metrics to assess their effectiveness, computational efficiency, and reliability. Results indicate that the HJaya-GA achieves effectiveness levels of 99.4%, 97.4%, and 97.1% across the three networks, while requiring at least 44% fewer function evaluations compared with the best-known metaheuristic results. Conversely, the Jaya algorithm exhibits marked performance degradation as the network scale increases. Methodologically, these findings demonstrate that simple metaheuristics can be systematically enhanced through informed initialisation and controlled diversification, providing a scalable and computationally efficient framework for large-scale WDNs design optimisation under limited computational resources.</p>