Instead of working with big monolithic caches that include everything from geometry to animation, you can work in layers. Working in layers gives you a super flexible workflow where different caches can be freely recomposed later on. This way you can use a static geometry cache during lookdev and merge it seamlessly with a deformed cache later during lighting.
Layering composition is directly performed by Alembic and USD libraries. Clarisse feeds the different file paths along their orders to the library and retrieve the result of the composition. Any performance loss resulting from layering is not coming from Clarisse as unfortunately, each time a layering operation is performed, these library reloads all the layers to perform the final composition.
In the following example, we exported a static geometry cache for lookdev, an animated one and a new static cache defining a new set of UVs. All these caches can be recomposed seamlessly in any order as if it was in the end a single cache.
Asset in rest pose with UVs
Animated version of the asset, exported without UVs. Notice how UV properties are still inherited from the rest pose
A new version of the rest pose is exported with a new set of UVs
Layers are reordered
The final version is the animated asset with the new set of UVs