"Dry contact" has nothing to do with "cleaning" or "arcing". It has to do with passivation mechanisms and Fermi levels. Physics tells, that with each metal to metal contact there's a certain activation energy needed to actually make the contact. This has nothing to do with cleaning from dirt or arcing, but is quantum mechanics.
It's a well proven fact, that only very certain materials can handle dry loads, which are called dry, because from the signal to be switched only very little or even no activation energy can be used to make the contact.
Once I had to repair a stage amplifier, where someone has exchanged a damaged output signal relay made of massive silver contacts by a relay with robust CdO-contacts. Although totally new this relay could not let pass small audio signals! Only when turning-on the volume and when loud signals came, music could be heard.
A typical (but not the only!) passivation process is the oxidation of contact materials by oxygen or sulphur. Then, rather complicated tunneling processes have to take place to make the contact. Tunneling is unlinear and results in strange hysteresis effects.


Normally, the only reason to put a gold plating on the contacts is to prevent the contacts from corrosion during shipping and storing. This gold plating is only a few µm "thick" and is burned away very soon after some switchings.
Only massive gold contacts sitting in an inert gas chamber can handle totally dry loads.


When using carbon pills every time you press the switch a small portion of carbon material of contact surface is wiped away. So, every time you press the switch you have a fresh surface. Unfortunately, this erosion of material results in a rather short life time of this type of switch.


Not at all. At least not for the standard switch contact of a keyboard. Just the opposite: It's a rather high current leading to a shortening of life expectancy. A good current to be switched with a standard keyboard switch contact is about 500µA...1mA. This current will show the least contact problems. In 5V systems this corresponds to a pull-up of 4k7...10kOhm.
But you are right, with much robuster switches 10mA can indeed be a dry load!

Kai