<p>Multi-agent systems encounter challenges from unbounded faults and delays due to their decentralized nature across networks. These faults, such as broadcast storms, can induce instability, necessitating the design of fault-tolerant systems with minimal conservatism to address induction delays. Recent studies suggest that delay partitioning and convex combinations can mitigate conservatism in stability conditions. Further research is essential to tackle unbounded faults and uncertainties in networked multi-agent systems by integrating delay conservatism techniques to enhance stability. This study introduces a new approach for fault estimation (FE) and fault-tolerant controller (FTC) to manage unbounded faults and unknown leader control inputs while minimizing delay conservatism and boosting system performance and resilience. The algorithm employs a delay-dependent augmented observer dynamic to estimate each agent’s state, infinite fault, and finite-order time derivatives of the fault. Fault dynamics depend on the delayed output signals from neighboring agents. This paper outlines an FTC and FE control protocol utilizing a cooperative controller to follow the leader’s trajectory and ensure agent consensus while accounting for interval time-varying delays as network constraints. The algorithm incorporates a self-accommodation term to counteract unbounded faults and a neural network-based dynamic that acts as a parameterized term for the leader’s unknown inputs. Pole placement is used to determine suitable gains for the FE and FTC algorithms. By comparing the overshoot ratio reduction factor of estimation faults to previous works, it is evident that the average ratio has decreased by 45%. Simulation results demonstrate the algorithm’s capability to tolerate and estimate unbounded fault signals, ensuring stability despite fault and delay constraints. Comparisons with alternative methods validate its effectiveness.</p>