I have been investigating a production problem for a circuit that uses several TLE2027 Op Amps. One of the parts in our circuit occasionally will fail on a brand new board because the output latches itself near the negative voltage rail during power-up and remains there as long as power is applied. The output snaps to the bad state at the point when the power rails have ramped up to approximately +/-3 Volts, a few hundred microseconds after power starts to turn On. If they remove-and-replace the "bad" part then it normally permanently prevents the problem from occurring again. The particular part that has this problem is the only one in our circuit that has zero impedance between the output and inverting input (it's in a Sallen-Key low pass filter). Apparently the problem occurs when the overall situation causes the power rails to ramp up slightly more slowly than usual. I found that I could fix the problem by inserting a 2K resistor in the feedback path. That seems to limit current out of the inverting input and prevents the problem.
So my question is: Is this behavior caused by current flowing through an incorrect internal path inside the TLE2027 when the power rails are rising slowly, and therefore causing an SCR to be formed that causes the output to latch? The implication is that there is a maximum ramp time requirement on the power rails (or a minimum impedance in the feedback path) that are not documented in the datasheet.