What is the cause of a shift in the FLD signal during a run?

That looks unusual, especially the peak with the flat top. In the Chemstation logbook there would be an ADC overflow error, I reckon? Not taking the baseline drop into consideration, the peak at 22min seems to saturate the detector. Remember, the highest peak has to be between 0 and 100 LU.

I would run an intensity test, the signal/noise test and the dark current test. A wavelength accuracy test should be done after  thorough flushing with water at 2 ml/min for about 15 mins. Pulling the diagnostic buffers and a system report might be required later. The tests itself are run with a restriction capillary installed and a flow of 0.25ml/min of fresh water.

Note that to be on the safe side we request to keep a distance between EX and EM of 40nm, so your parameters are borderline.

Was that drop also present on previous runs? I don't think so. As the baseline seems to follow the changes in the gradient, I suspect more an application issue, rather than a real hardware issue, but I would rule that out by running the tests. Still the flap top peak needs some attention. Also remember that well degassed, fresh solvents of very good quality are vital for FLD operation.

Have a word with an application chemist. I'm looking at it from a hardware side.

This question has been marked as assumed answered.