that is interesting because why would you use 100mA when you could put 30mA, except type S for distribution circuit.
It depends on what you are protecting, and how much it is expected to leak "normally" versus "faulty".
30mA is basically for human protection, though you get lower values.
But large systems, especially electronics, can leak more than that due to all of the capacitance L-E so you might need 100mA, or more, to provide protection against 'soft' faults that are not your MCB Zs-level of fault current (for example water getting in to a motor or junction box), or if the supply Zs is too high (classic TT case). Often such 'soft' faults quickly develop in to hard faults that can damage other things when they do go bang.
If you need selectivity then you have to have both a delay aspect
and a 3:1 or so difference in trip current. For example, a 100mA device could trip anywhere between 50-100mA and you want some margin over a 30mA one, etc, for small persistent fault currents.
But if you get a hard L-E short then
all cascaded RCD will have their current trip thresholds exceeded, so you need them to fire is sequence so the one closest to the fault fires before the one above it decided to act, etc. Much the same idea as cascaded OCPD really.