I think the simple answer is YES. Unless for example you have a high residual at a local point or region perhaps due to poor quality mesh or an unsteady flow for example. if you judge that the flow in this local area is not affecting your overall result you maybe able to ignore this local residual with caution. You can plot the residuals in a contour plot or volume plot to find out where the high values are and see if you can improve your model or mesh in that region to bring the residuals down.

Also during the solution you should look at the imbalances. For a steady state model you should set a convergence control on the imbalances. This will ensure that the flow in is equal to the flow out and that heat in is equal to flow out. The default settings do not have this convergence control activated.

In CFX you have the advantage (over e.g. Fluent), that you can easily switch between RMS and MAX residual plots in the CFX Solver Manager. If your residuals stagnate for some reason, than you can compare the two residuals for the not converging property. Rule-of-thumb: if the Max Residual is e.g. 2-3 oders of magnitude larger than the RMS Residual of the same quantity/variable, than this is a sign, that this maximal residual value occurs in your domain in only small number of cells/volumes/locations.

Further you can write the residual values to the RES file by checking the appropriate option in the output controls of CFX. After the simulation you will find signed values of residual values (not normalized !!) in the RES file and so you cn plot e.g. isosurfaces of the absolute value of the residuals. By that you can examine the extension and exact location of the regions with higher residuals in your geometry and fluid domain. With that both information you can follow the advice as given by Tristan Davenne.