I always say that transient noise is the option of last resort. The only reasons to use it are if the circuit is non-periodic in nature and can't be simulated with a more efficient method (such as semi-autonomous HB as described before) or if you need to be able to see a large signal response to the noise (e.g. the noise is large enough to cause distortion or significant intermodulation).
Transient noise is always a challenge, because you need to have sufficient accuracy to allow the noise (which is small for most circuits) to be observed given the usual dynamic range of transient analysis; low frequency noise needs a long simulation, and high frequency noise needs short time steps. So that's short time steps over a long simulation with high accuracy; that tends to mean long simulation times!
Small signal noise analyses are much more efficient and generally more accurate because you don't lose the noise below the numerical noise floor in the simulator, plus you don't need to run over multiple periods to see the average noise effect. So they're nearly always preferable if you can use them.