Two trip mechanisms in one plastic case
Open a domestic breaker and you find a switch, a spring, and two entirely separate devices whose only job is to let that spring go. The first is a bimetal strip: two metals bonded together with different expansion rates, heated by the current passing through them. Overload it and the strip bends, and after seconds or minutes it pushes the latch aside. Slow, and proportional — a 30 A breaker on a 45 A load will trip in a minute or two, on a 90 A load in a few seconds.
The second is a small solenoid with the current-carrying wire wound into it. At five to ten times rated current the magnetic field is strong enough to yank a plunger and release the same latch directly, in milliseconds. It does not wait for anything to heat up, because a short circuit is not a slow problem — it is hundreds of amps and it will destroy the cable in seconds.
Which means the diagnosis is already written in the behaviour. A breaker you can push up, that holds for a while and then drops, is being tripped by heat: an overload. Something is drawing more than the circuit is rated for, and the strip is doing exactly what it should. A breaker that clicks straight back down the instant you push it is being tripped by magnetism: a fault. Live to neutral, or live to earth, is finding a path it should not have. Reset is not the answer to either one; it is only the measurement.
Why the reset order matters as much as the reset
Here is the mistake almost everyone makes. They push the handle up with every appliance still switched on, it holds for a second and drops again, and they conclude the breaker is broken. It is not — they have just asked it to close into a live load. Motors draw five to seven times their running current for the first fraction of a second: an air conditioner, a fridge compressor, a washing machine, a pump. That inrush is enough to fire the magnetic trip on a circuit that is nowhere near overloaded once everything is running.
So the order is: switch off the heavy loads at the appliance, then reset, then bring them back one at a time. And if it drops again with everything off, stop after the second attempt. Every reset into a fault draws an arc across the contacts, and an arc pits the silver. Do that twenty times and the breaker will genuinely fail — it will feel loose, or it will hold but run warm, and then you have turned a free diagnosis into a part you have to buy.
The third device, and why "nothing was overloaded" is not a contradiction
An RCD — the wider device with the test button — does not measure current at all. It measures the difference between the current going out on the live and coming back on the neutral. In a healthy circuit those are identical and the difference is zero. If 30 milliamps is escaping to earth — through a damaged cable, a damp socket, a failing heating element, or a person — the difference is 30 milliamps and it trips in under a second. Thirty milliamps is not an overload. It is a tenth of a light bulb. So a house where nothing is drawing too much current can still trip its RCD all day, and it is not being oversensitive: something is leaking.
That is why the two faults need different searches. For an overload you are looking for what is running; for a leakage trip you are looking for what is connected, and often for one thing that is damp — a kettle that boiled over, an outdoor socket after rain, a washing machine whose door seal has been leaking into the motor for a year. Unplug everything, reset, then bring them back one at a time. The one that trips it is the one that leaks, and it will trip it every time, which is exactly what you want from a fault.