Normally I consider simple magnetic, low-wattage NPF choke preheat ballast about as reliable and safe as can be for lighting. Fortunately I was home when one of my terrarium fixtures decided to short.

I started to smell something electrical burning in the house...after a frantic search I noticed one of the terrarium F15T8 fixtures was not lit. It's a stock plastic Lights of America Unit I've had in daily use for over a decade, timer switched so not 24/7.

The ballast wasn't charred but was shedding insulation and extremely hot, starting to melt the plastic fixture housing. It appears to have shorted, and the tube had obliterated cathodes at both ends. Everything was working fine until it wasn't...it gave no indications of failure, but I wasn't in the room when whatever happened happened.

Any ideas on the cause of failure? Would internal fusing be a good idea? I've done this with a few old F40 Tulamp preheat ballasts. For a 120V 15w preheat choke, what rating of fuse should I use?

I've fused some of my small 20w preheat chokes, namely in fixtures that I tend to run unattended.

I checked the fuse and I have a 750 mA fuse in one of mine for a ~360 mA draw choke. One shot thermal fuses (available at electronic supply places) are an another alternative or can be used in tandem to a current fuse although you'll have to look up what temperature cut off you want the fuse to be at.

I think the thermal cutout is the way to go. Select one about 20degC above the maximum winding temperature rating.
And optionally you may add a manual reset thermal cutout as well, set at the rated max winding temperature (so it trips before the nonresettable type does).
In your case the culprit was likely the humidity above the aquarium, in my experience it feeds the rust like heck. Mainly if you are using those aquarium water heaters - they makethe water warmer than surrounding, causing the water to condense on anything above it, include the lighting components (when the lights are off).
Rust causes the core to expand, tearing the winding apart and so pinching the enamel insulation.

Medved, I think you are absolutely right. I noticed the ballast was quite rusty, so I had wondered if perhaps it became wet and shorted. What I didn’t know is that the rust causes the core to expand and lead to a short....so now this failure makes sense. It seems like a solution could be to mount the ballasts remotely where they would be away from the humidity.

As the F-can ballasts have a thermal cutout, it would make sense to add one to these ballasts too, or replace the ballasts with enclosed, thermally protected ones (which are hard to find in this size).

I believe the Advance LC-14-20-C-TP should have integrated thermal protection.

Would it be possible to remote-ballast the lamps..? Then the ballasts wouldn't be in the line of fire, as it were