I have never heard NI comment directly on this, but there are a couple things that would make me question the fine details of how common the execution order really is between highlighted and normal speed. To begin with, remember that the VI you are looking at may be highlighted but the subVIs aren't, so there can be differences in the relative speed of code - take loops for example.
Say you have two loops, one of which contains a single subVI and another that contains 4 or 5 low level primatives. Now let's say that when executing normally the subVI takes longer to execute than the low-level primatives. In this situation when running normally, (and assuming that the loops run the same number of times) the loop containing the primatives will finish first. Which in turn means that any further code that is waiting for the output from that loop will start executing before the other loop finishes.
However, when you turn on highlighting the loop containing the single subVI will execute quicker due to all the neat little bubbles running around showing the data flowing through the wires. Therefore the relative timing between different parts of code has shifted.
This is why you can see race conditions either become manifest, or disappear when highlighting is turned on.
Mike...
