Introduction
How a Sample Rate Error Occurs.
The Effects of a Sample Rate Error. (This doc)
Limitations.
How Ratefix works
The following example shows typical glitches that occur in a file output from one device at 44,100samples per second and read by the next device at 48,000 samples per second.
Every 12 or 13 samples a sample is duplicated as shown by the arrows.
The repetition rate is 
So in this example most repetitions will be after 12 samples but about 30% of them will be after 13 samples.
The sound of Sample Rate Glitches will depend very much on the Input and Output rates involved.
In the case of a 44.1kHz signal sampled at 48kHz there will be a hard edgy sound. The glitches (every 12 samples) will be introducing a component at around 3.9kHz.
The file will just sound wrong. On a fairly pure tone it will be very obvious but with more complex material it may not sound so wrong that it will be immediately identified as faulty by an untrained ear.
It is possible that you wouldn't even hear glitches introduced when the sample rates are very close. For example a file using the Pull Down rate of 47,952 sampled at 48,000 will have a repeated sample only every 1,000 samples. If however the glitch causes a Spike, rather than a repeated sample, it will be very evident as a click.
Sometimes when the receiving device clock is at a different rate from the sending device clock, the receiving device will read wildly wrong data instead of a repeated value. This shows up as a spike in the data (either positive or negative) where a single sample value is massively different from the samples around it.
Of course such 'spikes' can occur in valid data, particularly in synthesized test files, but they are much less likely to occur in normal music or speech.
Ratefix has a provision for removing such spikes when it creates a new output file.