On occasion it is not possible to refine the peak shape, ADP’s (Uiso values) and background together. This is usually in materials that have many peaks, and where the peaks start to overlap so that beyond some value of Q there are no areas without reflections so the intensities are always above the baseline for a large region of the pattern. These areas without peaks define where the background must be located. This means that if the high-Q background were to be placed at a slightly lower level, a comparable fit can be obtained by increasing the intensity of the high-Q data (by lowering the Uiso values for the atoms) and adjusting the peak shapes to put more intensity into the peak tails. The symptom for this will be a refinement that seems quite good in quality, but has very low, often negative, Uiso values. If a single Uiso value is negative and all other atoms have expected values, this can mean that the atom type is wrong or the occupancy is too low and more scattering density is needed at that site, but when all Uiso values are low this is an indication that there is not a single model to fit the data, but rather a range of values that produce similar fits.
My solution for this is to set all the Uiso values to a value somewhat larger than I would expect based on my experience for the material type and data collection temperature and then fit the background and sample broadening. I then stop fitting the background but do fit the Uiso values again. At this point the Uiso values will typically refine to larger values. Do note in any publications that the background could not be refined in the presence of other parameters and needed to be fixed.