While a number of people, (myself included,) have mentioned viral infection of mitochondria, I want to make it clear this is not as straight-forward a question as it may seem. Most of the time you don't find a virus infecting mitochondria. Murine Leukemia Virus itself is the one exception I've heard about, in mice.
The general rule has become almost dogma in some minds. This is because most of the time it is true. As anyone with ME/CFS should know, when exceptions occur they can be very important. The medical school rule of thumb "When you hear hoofbeats, think horses, not zebras." is handy for overly imaginative students, but researchers need to notice things like donkeys, mules, zebras and even the occasional okapi.
Here's my inexpert opinion on why you don't commonly find viral infections of mitochondria:
1) Many virions are physically too large. (You likely can't park a tractor/trailer in your garage.)
2) Even if the virion will fit, the entire viral genome may be too large and complicated to fit in a plasmid mitochondria can cope with. The limit seems to be something less than 10,000 base pairs, perhaps near 8,000. There are relatively few viruses with genomes that small.
3) Even if the virion and viral genome will fit, the process of converting the information into molecules which form the virion may require more sophisticated support than is available inside a mitochondrion.
This last point sounds obscure, but it can be illustrated with a gotcha in common modern computer technology. Consider the cellular machinery which translates DNA as an operating system. For the whole cell, we have something like Windows 7 (tr), with support for all kinds of obscure things needed for compatibility with older versions. For the stripped-down machinery of mitochondrial transcription, we have something like the OS of a Palm Treo (tr). If you have tried to run a Windows (tr) application on a Palm Treo (tr), you might guess a few support libraries are missing. (I could go further and compare the complexity of a complete cell to what system programmers call "DLL Hell". This, however, takes us away from the central point here.)
The bottom line is, yes, it could be true; no, it doesn't happen often or easily.
Real researchers will have to take over from here.