This paper examines the impact of faulty nodes on Practical Byzantine Fault Tolerance (PBFT) algorithms, focusing on the AWARE optimization. While AWARE improves average-case latency by assigning larger voting weights to well-connected nodes, it is vulnerable to exploitation by Byzantine nodes. We propose modifications to AWARE to enhance resilience of dynamic link latency estimation. Experiments show that comparing predicted and actual consensus latencies helps detect and mitigate performance issues caused by malicious nodes. Integrating latency information with BFT-SMaRt’s leader change algorithm offers a more robust solution. We also analyse shortcomings of AWARE dynamic leader selection system.