The successive interference cancellation tree algorithm with first success (SICTA/FS) was specifically designed as a random access protocol over noisy collision channels. Given J users with an infinite buffer and subject to Poisson arrivals, SICTA/FS achieves throughputs as high as 0.6 if packet losses are allowed (up to 20%), while without packet losses its throughput quickly degrades as the number of users J increases. In this paper we indicate that SICTA/FS may remain stable for a considerable amount of time before becoming unstable when the arrival rate exceeds the maximum stable throughput. More importantly, we also study the ALOHA-SICTA/FS algorithm and show that it can achieve throughputs of 0.6 or above without packet loss. Additionally, we provide an accurate estimation of the mean packet delay under ALOHA-SICTA/FS using a simple queueing model with vacations. Finally, we indicate that ALOHASICTA/FS suffers from hardly any throughput reduction when the access point stores the last two collision signals only.