It's not all soft. And sandstone isn't really all that hard, either.
The point is -- with enough water energy (volume and speed), and enough abrasives, you can dig a pretty deep hole pretty fast. As an example, the Coulee area in Washington State was formed in hard basalt over a fairly short interval. (See, for example, this article on Dry Falls, which was formed over the space of 2,000 years when a huge ice dam ruptured.)
If you want an impressive feat of hydrological digging, go to the Black Canyon of the Gunnison river. Wow!
How long does it take for sandsone to form. From personal experience with wells, about 2 years.
Also, the puzzle of the GC is that it cuts across and incredibly large alluvial plain, (ie... the north side is higher than the south.) There is a large amount of evidence that this was formed rapidly. Also, there seems to be a surprising amount of evidence gathered by the USGS that it was formed by a large inland lake bank failure and was created in as little as three months to a year. The "official" USGS story is still the old one though. What hurts the new theory is that it is applied by some to confirm the Noah's flood. Anyway, the GC may not be all that old (geologically speaking.) Or, if old, could have been formed instantaneously on a geological timescale.