By characterizing the worst case profile, which maximizes the content of a buffer fed with leaky bucket regulated flows in packet telecommunication networks, we derive a tight upper bound in the many-sources regime for the tail distribution of the workload generated by these flows in a FIFO queue with constant service rate. Furthermore, we compare this workload distribution with an M/G/1 queue and get insights on the better-than-Poisson property of regulated flows. We conclude that the superposition of independent regulated flows generates an asymptotically smaller workload than a marked Poisson process whose service times and intensity depend on the parameters of regulated sources. Key words: Network Calculus; Statistical multiplexing; Better than Poisson