The short answer is probably not.
See this article for a short description of how these things are done.
For coins, the machine measures diameter, thickness and number of ridges (on part of the circumference - not the whole edge) so you may think that grinding the coins would affect this.
However, because coins are treated pretty harshly - banging against each other in our pockets and purses, the machines have reasonably high tolerance bands and a coin has to be substantially damaged before it will be rejected in the first place. The whole point is to recognise the denomination and that the coin is domestic. Most coins worldwide are different enough that even with wide tolerance bands foreign coins get screened out. Australian and New Zealand 20c coins being a notable exception.
A more likely explanation is that the coins are fed in too fast for the machine to process and they will shunt the excess to the reject chute. Rubbing the coin will simply give the machine time to clear itself. So would counting to ten!
The answer above has, quite rightly, been criticized so I leave it here as a lesson to myself on what not to do. I will now approach this more rigorously.
How does a coin detector work
I have been unable to determine the exact model of detector used in rapid transit ticket machines in Germany. This indicates that there are 5 separate U-bahn operators and a single S-bahn operator - it is possible they all use the same detector, or maybe they don't, or maybe there are different detectors in use within each system.
This is a modern coin detector. According to the literature it tests "40 measurements of coin parameters", for obvious reasons it doesn't enumerate them. Other products on the site check fewer parameters.
Notwithstanding, from this general overview, we know that some of the parameters measured include:
- optical characteristics (diameter, thickness, ridges, head & tail embossing)
- alloy detection (through use of the Hall Effect)
Coins that were initially rejected are, after being scratched on a metallic surface (either on the face or edge), more likely to be accepted on subsequent presentation than coins that were not scratched.
Evidence in support of the Hypothesis
None (anecdotes are not evidence).
Possible reasons why it may work
Given the way coin detectors work, none.
Scratching a coin will not appreciably alter its mass, size and shape or the alloy it is made of.
It may remove some surface dirt but probably not enough to affect its optical signature.
It may damage the coin but this is more likely to increase rejection rates.
It may remove or add static electricity, but none of the detection methods would be affected by this.
Given the lack of evidence for the phenomenon and the lack of any physical mechanism by which it can occur we reject the hypothesis in favor of the alternative hypothesis - it doesn't do anything.
That is not to say that people don't do the coin rubbing thing, at least in Europe - the blog article the OP cites is evidence of this. However, here in Australia I have neither heard of or witnessed the practice - a statement that you should disregard as mere anecdote.
This means that the alternative hypothesis comes with a corollary - people are idiots.