When (I2C) clock stretching occurs, its the slave doing this, not the master. See the links Danan provided. So basically this is something I2C-specific - the slave can signal to the master that it wants to go slower for some reason.
But actually something similar can occur with other protocols like SPI - but there the master needs to stretch the clock since the slave cannot influence it. (And I have not seen it there in real life yet)