I could, legally make 5% resistors guaranteed from -35 to +65 degrees with no published tempco which could be as follows:

a 1000 ohm resistor would be 950 ohms at -35 and 1050 ohms at +65 equivalent to a tempco of .1%/degree C

With such a resistor 'measured' to be within .1% it would fail if the temperature in the actual circuit was 1 degree different from the measuring.

A dramatic example, perhaps, but how else to get it through is skull

Erik