Importance of removing ripples in output of dc-dc converter under various conditions

I was just wondering. I heard from somewhere that for some power supplies, it is important that there be no ripple and 'dips' in the output of a constant voltage output dc-dc converter. The specific cases are:

1.) When dynamic loading or hot swap testing, basically transient conditions it is not desirable for the output to have small negative dips or some oscillations even though they will eventually regulate at a fix value at steady-state.
2.) When the power supply is turned on, the output voltage must be monotonically increasing, if there is an upward spike which is small that is okay as long as there is no downward spike.

I heard that it depends on the load but is this really that important? Are there examples of load you could give which are really sensitive to these kinds of specs?

Also, I was wondering what if, if you apply a step increase in load, the output voltage of 5V dropped a little bit let's say 0.05V, becoming 4.95V and stayed there but it did not oscillate and it regulated there. Would that be better than if the step increase in load caused a downward dip of 0.05 and possible oscillation but in the end after some time to reduce the oscillation it regulated to 5V and not 4.95V. Which is better?

A quick review of the reset condition for a few MCUs such as the SiLabs C8051Xxxx, Ti LM3S9D90 or a common supervisor device such as a MAXIM MAX815 will give you a range of thresholds where devices will automatically reset if the supply voltage drops below threshold. They tend to be 2/3 of their operating voltage. In the case of the listed 3.3v devices this is around 2.1v. The noise including ripple on a typical 5v bus is around +/-0.05v. If you maintain the voltage at +/-10% with a step current change everything should be ok.

There are special devices to include high frequency RF and optical devices in GaAs and InGaAs that may be damaged by even a small spike <0v or a small overshoot in excess of Vdd.

Also some instrument applications where very low noise is required would required very low ripple but these types of circuits are typically isolated from power supplies subject to hot swapping or other transients.

The output-monotonicity requirement is really outside
the converter's ability to determine, unless it has a
perfect output voltage stiffnesss relative to the load
activity - such as if the driven load has its own UVLO
threshold and starts abruptly throwing power, the loop
will only react later to the sudden draw and there's
your negative dip.

Now significant negative dips stand a chance of hanging
up the turnon cycle, because that same UVLO in the
load might then reset the load, output comes back up,
UVLO releases and you just repeat indefinitely or chatter
(potentially causing some sort of overstress) until the
output has risen past the other side of the deadband.

You write first in the opening "I heard from somewhere that for some power supplies" .Do you have any link from where you find that devices???

Thanks for the reply guys! Now I understand how that negative dip becomes a problem. About the oscillations that occur during step load changes, is there any problem associated with it? Let's say that the dip that occurred is not that big enough to cause problems related to UVLO and reset but there is still an oscillation due to the step load but it dies after some time and regulated to a fixed voltage value is there some problem that it may still cause, or is this okay?

About "I heard from somewhere that for some power supplies...", I heard it from engineers talking and I didn't see it online, I just forgot to ask them about it before.

This is the internal schematic of the meter. Resistance means nothing as I used it to connect the diagram. Here I got two element but I DON'T know which one is CT and what other did in the meter. As far as I think, one would be used for the Neutral missing and one is for the isolation purpose. Small one in for the isolation purpose and the other would used for the neutral missing tamper case. Correct me if I am wrong. So now I want to know that I want to design a meter having two shunt or two CT. Is It Possible to design a meter????