No, but I've seen it in a Chipmunk!
There will actually be a sequence of roll and pitch errors in a full 360 degree turn; steep or not. In the Chippie (and many other designs of that vintage) the A/H is suction powered and the base of the main rotor case is slightly bottom-heavy. This pendulous mass is acellerated by the centripetal force during the turn and therefore the bottom of the rotor housing 'swings out' very slightly. The force causing this apparent bank error is then precessed by the gyro to cause an actual pitch error. As the turn progresses, the combination of centripetal acelleration and precession causes a combination of pitch and roll errors that reach minima and maxima every 90 degrees of heading change. When the turn is completed and the wings are level, the pendulous vanes quickly remove any residual error.
I was once told that the top of the gyro was displaced very slightly from the vertical in order to minimise these effects but this was all about 37 years ago so the memory is fading somewhat!
The pendulous vanes themselves are also subject to acceleration errors of course.
Not sure how this applies to an electrically driven gyro as fitted to the B737-200.
Merry Christmas!