I am running a gradient method (30% Acetonitrile to 50% Acetonitrile in 30 minutes) on HP 1090 HPLCs. I have noticed that (sometimes) the relative retention times of impurity peaks relative to the active peak are not consistent from one day to the next. Is this a common phenomenon with gradient methods? Could it be from the plumbing in the HPLCs themselves?

Are they inconsistent from one day to the next or from one system to the next?

This is not a common phenomenon in a well developed gradient method. There should be some definable cause.

How sure are you that your pump is working properly? Imperfect mixing or total flow rate could easily result in the problem you note.
You can check flow by timing the collection of the HPLC effluent in volumetric glassware.
You can check the consistency of the mixing apparatus by spiking your 50% acetonitrile with a solvent with a chromophore (acetone or toluene are often used) and checking the linearity and consistency of the resulting (sloping)"baseline".

How sure are you of the goodness of your column? Swapping columns might be worthwhile.

Is there anything in your sample that could be affecting the chromatography -- large amounts of excipients without chromophores? Or, are the differences in retention times in any way related to changes in the total quantity of material injected? You can coat parts of the column temporarily, and affect the way other components are retained. This would be especially true for components eluting after the large components (visible or not).

Is there any overall trend to the change, AND, is there any chance that not all the material you inject is eluting? Try running a gradient to much higher levels of acetonitrile (say, 95%) and see whether anything else elutes. (If there are late-eluting components without chromophores, you may have to use secondary effects to see whether such a wash-out helps.)

Good luck.

I don't have any experience with the HP 1090 but if it uses a solenoid operated mixing valve on the low pressure side of the pump to form the gradient you might want to check/clean/replace this. This problem has come up a couple of times with our HP 1100s.

Good Luck.