I would do all the compositing ahead of time, then interpolate the final piece up to the final size. The main reason is because the file size of a 40×89" image full of layers would be huge. You’ll also (I’m guessing) want to make smaller prints for proofing purposes before the final design is ok’ed.
Now that being said, it seems that 150 ppi is excessive for a banner that size. Large prints like that are not intended to have perfect detail because they are meant to be viewed from a distance. If that’s what they asked for, I would question them, but maybe that’s really what they need.
Actually the piece is for a 3D art installation and is made up of seven of these panels running about 20′ in length and, unlike billboards, the viewers are going to be able to come much closer so the higher dpi is necessary and even then we’ll have to live with a softer image even after some sharpening which poses it’s own problems because normally the sharpening on something this big could be more aggressive due to a larger viewing distance.
I would not bother to interpolate the images since the Lambda will do a fine job of it (I made comparative tests back when I was operating a Lambda and you can’t tell the difference between PS Bicubic interpolation and the Lambda’s). Just lay everything out in Quark or ID so that the type and other vector elements are nice and sharp
If you prefer to interpolate anyways, then do so, but I’d recommend 200 ppi. The reason being the Lambda will have to interpolate the image to 200 ppi when outputing.
Also, for a mural that size an image resolution of 72 to 100 ppi is usually plenty since most people will not view the image up close (there are exceptions to this rule of thumb)
Finally, talk to the folks with the Lambda, they can probably give you some good pointers. (and remember, keep the file in RGB)
Use Smart Objects placed in your full res file so you are not resampling anything until your final flattened version. You will probably find editing a lot quicker this way, and a lot of your finer point editing can be done on the original low res data. And also check what the optimum resolution is before deciding what the res of the full file will be. It no good having the lambda upsample the file.
Jpeg compressed EPS will allow you to get away with huge files at a much smaller file size (maximum quality jpeg and theres practically no quality loss).
