You are correct, the limit is the overall current budget, not due to Q16 alone. But from a reliability standpoint, it helps that the 6030 has a max Tj of 150 c (that's pretty close to softening 63/37 solder 
. Also, I saw something about continuous Ids of 10 A at 115 c. So we're very safe, even if water boils off it in normal operation. However, the glue that binds the laminate and the copper to the laminate may deteriorate at higher temperatures, and some of the dissipated heat may affect adjacent components.
However, as you rightly observed, there is sufficient margin with a case temperature of 50 c. My observation is that 4x66 at 2.0 v is well within the power budget, provided the hard drive does not load it over the power budget (of course, the HD runs off the 5v supply).
For a dissipation of 10 W, they should have seriously considered a switching regulator, similar to the designs used in Voltage Regulator Modules (VRMs). Those 10 W would have gone a long way in helping to power attached USB devices.
While we're on the topic of temperature, the other issue is that the K6-III uses the IBM C4 flip-chip solder-ball process to bond the chip to the ceramic substrate. On the plus side, the max current-density of the solder balls is higher than wire-bonded pads on Intel chips. So it is much more resistant to failure due to high current densities. The balls themselves can soften at higher temperatures, but that is not particularly important, since it will not come into play until we get to 200 c or higher, by which point the chip would anyway have died due to other unnatural causes. In short, the C4 process packages can withstand stresses due to overclocking and improper cooling somewhat better than traditional wire-bonding.
