Delay-time sensitivities for different η and different threshold positions. a, η = 1/2. It can be seen that for all threshold positions the sensitivity of the delay time is smallest for non-linear decay and largest for accumulation. b, η = 2. Also here the sensitivity of the delay time is smallest for non-linear decay and largest for accumulation for most threshold positions. Note that the situation is reversed for high threshold levels corresponding to high sensitivity in all cases (fig. 1). c-e, Delay time sensitivity as a function of η and PT for the cases of accumulation (c), decay (d) and non-linear decay (e). The logarithm of the delay time sensitivity is shown: log (|η - 1|/|δt|) for decay and log (|η-1 - 1|/|δt|) for accumulation. δ t was normalized by η-1 for decay and by η-1-1 for accumulation, which correspond to δ t in the non-buffered case in which T1 = η T0 for decay and T1 = η-1T0 for accumulation. Thus, blue represents a non-buffered system. Red represents a buffered system.