One of the more interesting things I did in my career was operating copy cameras for several years at National Geographic in the mid-1970s. One of the cameras filled an entire room and had a 6 x 8 foot copy board and a 30 x 40 inch film back. From National Geographic I moved on to managing printing where I had occasion to color separate and print art catalogs from customer supplied transparencies. As such I'm aware of process control factors beyond just capturing the image that need to be addressed when shooting art to ensure the most accurate possible reproduction.

The standard copy camera set up is lights at 45 degrees to the glass copy board. When the lights are angled that way the reflections of the lights are not seen from 0 degrees on the camera axis. Fire a flash angled at 45 degrees to a mirror with the camera square to the mirror and you'll see what I mean. At 45 degrees the flash will light the mirror but the reflection will not be seen by the camera oriented at 0 degrees. So while polarization certainly will not hurt, it should not be necessary for flat art such as watercolors on matte paper if you set your lights correctly. We routinely made halftones and color separations from glossy photos on glass copy boards without any reflection problems.

Where cross polarization is needed is when photographing irregular surfaces which reflect the light off the side at angles which cause the camera to see the specular reflections: an acrylic or oil painting with heavy textured brush stokes for example. Cross polarization can kill some or all of the reflections. Kill all of the reflections and the impression of texture in the painting will be lost. Mostly it depends on how the photos will be used. When in doubt shoot both ways.

Don't shoot the art angled on an easel. Find a blank wall in a room where you can turn off all the lights. Your biggest problem with reflections will not be the lights you use to shoot the art but other sources hitting the glass and reflecting into the camera.

Here's a simple set-up to ensure the art is square to the camera: Measure off a square on the wall bigger than the largest painting with the center at the height of your camera on the tripod. With painters (non stick) masking tape create an X on the diagonals. Put small mirror over the center of the X. With a bubble level in the hot shoe or tripod QR, level the camera then adjust until you can see the reflection of the camera lens in the mirror. You might be able to see your eye in the viewfinder. That will put your camera square to the wall with the center of the X on the lens axis. Then all you need to do is use the tape as a guide for the corners of the artwork to keep them centered on the lens axis and square relative to the camera. Set the lights for even lighting on largest painting (more on this later). You will only need to do the set-up once to shoot all the art and the exposures of all will match.

Other key concerns are even exposure and color fidelity. We insisted that all photos of artwork contain a Kodak Q-13 color separation guide. It consists of the grayscale on the right and color target printed using offset process inks on the bottom of this lighting test shot below. An even better color reference today is the MacBeth ColorChecker, the target on top. Its colors are also within the gamut printing presses and ink jets can reproduce. The target on the left is a QP card. They come in packs and are relatively inexpensive and ideal for putting around the artwork to monitor exposure evenness.

The reason we insisted on color and tonal references in every photo is because a camera will not reproduce color accurately or consistently, but Kodak and MacBeth go to great expense to ensure every target they sell are is close to identical to each other as possible. That is what allows them to be used a a standard reference for color: every one sees the same thing and knows whether or not it is reproduced accurate or not.

This is the Macbeth Chart, per the RGB values in the documentation, recreated in Photoshop by the numbers:

Here is the actual chart, shot with my 20D in RAW with styles applied in DPP:

For a "pleasing color" shot I'd try them all and pick the one which I liked the most for that particular photo. Reproducing art is a bit more difficult. Often pigments used in artist colors are beyond the gamut a camera can record accurately or a printing ink can reproduce. Often the lighting used would not be neutral or processing of the transparencies would introduce a color cast. We might not know what the painting was supposed to look like, but we had copies of the same references and knew what CYMK values they should have when color separated and printed. So when separating and printing we aimed for reproducing the target as accurately as possible, letting the painting come along for the ride.

The best that can be expected in reproduction is accurate rendering colors inside the printing gamut. Colors outside the printer gamut will shift. Customers expect the reproduction to match the art and will be disappoint when comparing the original with the reproduction unless their expectations are lowered from "match the art" into the realm of the possible printers live in: matching the target. Just proofing the image can be problematical because the proof will differ from both the original and art the final output if produced on a different printer than the proof. Being able to show the client that the target was reproduced accurately at the various stages is a way to demonstrate that the reproduction is as close as physically possible.

When shooting RAW it is not critical to set Custom WB in the camera off a gray for the final output because it can be snapped to neutral from the target image in the shot, but it is important if using the in camera feedback because it will ensure the neutral test targets clip in all channels at the same time. It also helps perception when editing because the Custom WB tag in the file will ensure the color is neutral when first displayed on screen.

A simple way to determine if exposure is even is to use a target like the QPCard in each corner and one in the middle when setting the lights. Open the aperture until one of them clips. Adjust the lighting until all four corners and the middle clip at the same time and you will know the lights are even. Then close the aperture about 1/3 stop so the white patch reproduces with an eyedropper of 250, the correct value for a smooth solid white. The only 255 values in the file should be specular reflections, should you choose to render brush stroke texture.

How much of this applies to what you are doing will depend on what the images will be used for and how critical his requirements are. Its best to find that out up front. Since you are in uncharted water and seem to have no idea where the mines are (I've tried to reveal a few) you might want to do a test shot for approval before tackling whole project if it is a large one. Like weddings it is a job best left to specialists (who use copy cameras) if the client needs are critical.