You can render one frame and see what takes most time: the pre-rendering steps like parsing, calculating displacement, etc, or maybe the actual rendering process?
In case of no moving objects (fly-through) it's easy to save the GI precomputation time by reusing a single GI cache file for all frames. This way the GI does not have to be recomputed for each frame.
See:
https://support.chaos.com/hc/en-us/articles/4528617365649With moving objects/lights it is more tricky, but if they are moving smoothly, and not very fast, you might get away with using the "load+append" option. That can also help reduce GI calculation times per frame.
If you don't want to save your frames in a high bit depth format (like 32-bit EXR), you can go to the System tab and change Highlight Clamping (not highlight compression!) to non-zero. Values like 1-2 should work, but you may need to find the right one by trial and error. This will clamp the brightest pixels in the image so you will actually get better quality of things like hard edges and shorter render times. The drawback is only that it clamps those highest values so you lose the high dynamic range.
Simplifying materials helps with render times. For example: using a single Corona Physical material with texture masks for different properties rather than a Layered Material with two different materials in it that are mixed together (obviously this is not always possible).
It really depends what the scene looks like and what features you are using. If you could share some images/screenshots/sample frames, that would be very helpful.
You can contact the support team here:
https://support.chaos.com/hc/en-us/requests/new