The brain links bits of sound together when their relationships would be very improbable if they came from unrelated events. Sounds that share a common pitch base, fit a harmonic pattern, or come from the same place get grouped. This is how two people talking at once can be pulled apart.
Everyday listening hands the ear one jumbled wave made of many overlapping sources, and the brain must break it back into separate sounds. Even basic qualities like loudness, tone colour and location are only heard after this sorting, so they belong to the sorted-out sound, not to the raw wave.
A fast series of sounds alternating between dissimilar kinds is not heard in the order played. Instead it breaks into two or more streams heard at the same time, each holding similar sounds, much as similar shapes group together in vision. You cannot attend to two interwoven streams at once, which is why the order across them gets lost.
A piece of sensory evidence is normally assigned to only one perceived sound at a time, so claiming it for one source takes it away from the others. Rare exceptions, where the same sound feeds both a speech percept and a plain noise percept, may happen only when separate mental systems each build their own description.
When a louder sound joins an ongoing one, the brain assumes the old sound carries on underneath and treats only the extra energy as new. That is why a soft tone seems to continue straight through a loud burst of noise that actually replaced it.