One of my F15T8 undercabinet fixtures felt awfully hot.  I measured the temperature of the exterior near the ballast and it was 120 °F.  Does this suggest that the ballast is overheating?  What might cause it to run so hot?  My other fixtures are only moderately warm to the touch.

Was it always running hot ? If not, then it might be the ballast windings that are gradually shorting out. A ballast failure is probably not that far in the future.

Is it of the same brand as the other ones tha are running cooler ? It might be of a different design as some ballasts run hotter than others. There are even cases (documented in some posts here) where different models of the same manufacturer run hotter than others. Some members have reported that Vossloh Schwabe ballasts from France run much hotter than the ones from Germany, for instance.

Hope this helps.

I have some metallic F15T8 undercabinet fixtures laying around. Might be interesting to mount them all below a good size board, run them and take a heat reading. I do plan to eventually put some in my kitchen...so I would want the coolest running ones.

If you want to know if your ballast really overheating, you can measure its body temperature. A ballast temperature of up to 230*F (110*C) is noraml for most ballasts (When the core temperature is the same as the ballast max Tw value [248-266*F 120-130*C in most ballasts. Data given from Eltam website]). Above 230*F means that the ballast is overheating.

Here is quite important to note, than the US style potted ballasts have still quite large temperature difference between the ballast surface and the hottest point in the winding. So when the winding only 130degC and there is 60degC temperature difference, it mean the ballast surface temperature become "at the limit" already at 70degC.
The European style ballasts are usually quite naked (you directly see the main core,...), so the surface temperature differ only a little from the hottest point (~10degC is quite common), so for the same 130degC in the coil it mean the surface could be as hot as 120degC, yet not creating higher stress than the potted unit with only 70degC on it's surface.

But on each ballast is written the maximum allowed temperature "Tc". Either there is no reference point, that mean you could measure on any thermally conductive point (when the ballast have always uniform temperature across it's surface; the case of European magnetic ballasts), or there is specified a "reference point", where the temperature shall be measured.
Now to determine, whether the ballast is overheating or not, just measure the temperature there. But to be valid, the fixture have to be fully assembled. So you will need a thin thermocouple, or heat sensitive stickers (irreversibly change color, when their temperature exceed given limit) for that.

When measuring ballast surface temp, I scan the entire area until I find the "hot" spot...usually only about 2-3 cm square area. On average ballasts...that feel hot but not painful right away, it's usually around 120-130 F and when they get above 140, they start to really feel hot..then I reconsider either better heat dissipation (fixture location, fixture style, etc.) or a different ballast. Even if temp still falls within specs, if at the upper end of scale, ballast life will still be affected. Then other factors to consider are ambient temperatures which can drastically change ballast measurements. Also, surface mounted fixtures tend to trap more heat inside as well.

@don: With that you find only the surface temperature, but what matter for the life is the internal temperature.
And there your method could easily fool you, unless you make sure to really compare "apples with apples":
The European style ballast (so naked core, well impregnated winding) could run above 70..80degC on the surface, while it's winding hot spot would run at about 90degC, still way cooler than ~120degC inside the potted ballast designs with surface temperature only 50degC.
So even when the naked design run way hotter on the surface, it will degrade way slower than the colder surface running potted design.

So way better approach is to measure the designed test points and determine the temperature margin till the maximum rating defined for that spot: So the naked ballast running at 80degC have more 50degC margin towards the specified maximum of 130degC, but the potted design running at 50degC is only 10degC from it's maximum of 60degC.
You may assume all ballasts are designed for about the same failure rate and lifetime and the "10degC colder equal double life" rule apply there, so each 10degC you save from the rated temperature mean double the life, compare to the ballast rating (usually 100kHours)...

P.S. I used the Celsius units, as I'm familiar with those numbers, you can convert them to degF, but I guess it is not really necessary for understanding the concept...

@Medved, Yes, I'm well aware of only surface temp, but my measurements are only RELATIVE. In other words, after I take enough readings from different ballasts I start to see what a reasonable temperature appears to be to at least HOPE it will help me to select ballast, fixture design, etc. I'm not familiar with these "test points" you mention, but I would like to know. At any rate, assuming most ballasts are filled properly with tar, the same size cores should dissipate heat at a similar rate. As I said...it's just relative measurements...comparing to other ballasts. It's really nothing scientific. Just common sense.

I mean this potted style versus this "naked" style.

And with the tar filling is still one problem: Even when all potted could use same materials, the geometry may differ (distance from the hot core to the surface metal) and that could cause quite large differences in the thermal resistance between the inner workings and surface (although not as large as between the potted and naked).

For the temperature measurement the rule is to follow the test point specification (on this ballast the black dot bottom left from the "CE" mark; there may be multiple of such test points across the surface, you have to verify all of them) and if not explicitly specified, the search for the hottest point is indeed the correct method (I double checked some pictures of the US potted ballasts and I haven't seen such specification on the label, but there was no spec on the temperature either, so all that could be then still in the datasheet).

And I just remember one method, how to check the temperature of a winding, where you can not put a thermometer, so with ballast it would be in fact the temperature of the hottest part:
The winding is made (I would expect on all magnetic ballasts) of copper, what change it's resistance with temperature in a known manner (thermal coefficient you could find, I guess).
So if you measure the resistance of some winding (e.g. 120V primary, it is usually the best and fastest accessible one) after the ballast was sitting the day on your bench (so have the ambient temperature), then power the fixture for few hours (till it thermally stabilizes) and then just after disconnecting the power (before it cool down) measure the resistance again.
Then calculate the ratio of these two resistances, divide it by the thermal coefficient of the copper resistance and viola, you get a temperature difference between the primary winding and ambient. As we can expect the winding temperature would be the weakest point and then the used materials have similar resistance to high temperature, it could be a good starting point to determine how the ballast is stressed inside...