Since the two converter outputs are connected together, this can produce +5V from whichever supply is present. In other words, if the 12V supply is absent, then +5V is derived from the 3.3V supply, and vice versa.
D9 and D10 form a "diode-OR" configuration which "steers" the greater of the two converter outputs to the ultimate 5V output, and simultaneously prevents the output of either converter from injecting current into the other.
You might think that this would cause the output to drop under 5V (by 0.7V), but cleverly the designers have taken feedback from after the diode, to compensate for this.
Edit: I suspect that the designers also configured one of the converters to produce a slightly greater output voltage than the other (like 5.0V and 5.05V), so that when in operation, the other converter senses an "overvoltage", and ceases its switching, becoming effectively idle. This function is perhaps implemented in the large transistor network preceding both converters, though.