Exactly the same thing happened to me. It's fortunate I found a
service manual for my frequency counter on eBay or I might never
have diagnosed the problem. The manual states that the normal
operating temperature for the oscillator is 80-84C.

Henry-

I had the service manual for my counter, but it did not have sufficient
information about the 10811-series OCXO, for the HP/Agilent parts department to
identify the thermal fuse. I had to obtain a separate manual for slightly
different oscillators in that series.

The temperature of the oven is chosen to match each specific crystal. The
thermal fuse should be sufficiently high that it won't open in normal
operation, but will if there is a problem. The original part for my oscillator
was rated at 108 degrees C. The revised part is rated at 115 degrees C. It is
HP/Agilent part number 10811-80008. Last year the price was $10.50 plus tax.

I also found a 117 Degree C "Thermal Cut-Off", NTE Electronics NTE8115. Radio
Shack has a #270-1322A thermal fuse rated at 128 degrees C. Both of these are
larger than the original part, and their leads are too fat to fit in the
original pin sockets. I solved that by finding a high value resistor with the
correct lead diameter, and soldering it in parallel with the thermal fuse. It
was necessary to use a heavy pair of pliers as a heat sink to avoid melting the
fuse element while soldering! The larger thermal fuse with its parallel
resistor is a tight fit. I used some fiberglass tape to insulate it, to keep
from shorting to nearby components.

I eventually replaced the substitute with the correct part, and it has been
working for about a year now with less than a tenth of one Hertz drift at 10
MHz, or better than ten parts per billion compared to the Rubidium controlled
oscillator.

73, Fred, K4DII