Hi sssweb, as Lucy has said, this isn't a case of "splitting" a:hover because that isn't a single selector, but two combined:
a = element
:hover = pseudo-class that applies on hover
One way of thinking about this is that it makes sense to write the hard-coded element and class a.text (which don't change) before specifying the dynamic pseudo-class :hover (which does).
However keep in mind efficiency: Your selector requires the browser to evaluate every <a> to check if it is also classed .text. Unless you have lots of different link styles, more efficient to just use a:hover.
Note modern browsers accept :hover on elements other than <a>. More detail and the full range of pseudo-classes (and pseudo-elements) in the link to the recommendations given above. Also note css3 pseudo-elements use double colon syntax - so Lucy's example of :first-letter will become ::first-letter - but it will be sometime before that syntax is fully understood.
In terms of the SelectORacle instructions, I read that by reference to the first part of the sentence:
Best of all, the SelectORacle will flag potential errors and other problems, and it won't choke on any actual rules. (My emph)
.. while the tool flags errors, it can interpret complex selectors - so convoluted selectors won't be (falsely) reported as an error. But my thoughts only. As we recommend the tool so often I've emailed to ask. Edit to answer question better
Best I can tell SelectORacle definitely stops/reports syntax
errors such as misplaced punctuation and out-of-order pseudo's. It will explain selectors such as red:hover, although red isn't a valid HTML element - but it might be in xml - recall this tool is for explaining the selector - not checking HTML validity.