There is a denoising function (I don't remember the exact name), which will take care of many artifacts you may see.
You would normally enable it a few frames before the end of the render. You can see if that gives you improvements without doing so many iterations.
Also, there is a reduced rate of return, because every time after an update it will take more cycles to get to the next update.
For animations, the only realistic number would be less than 500 and use denoising.
You can also render at a smaller size and use something like Topaz Gigapixel Ai to enlarge the image. Sometimes that will give you more details/sharpness than you would get with just rendering at the larger format.