We have a use case with a very large (85+ GiB, 20+ layers) dataset with small features that only need to be served on one zoom level. "Small" means that the features are usually not bigger than 1 tile in the single zoomlevel that they are served on. The features in the dataset are validated beforehand (self intersecting, etc). Clipping, simplifying and cleaning/validating are not necessary in our case, and disabling them increases performance. Especially cleaning takes a lot of processing time, which takes the response time above a workable threshold for this dataset. (Especially when clipping is disabled, but that's a known issue, I believe.)
We have a use case with a very large (85+ GiB, 20+ layers) dataset with small features that only need to be served on one zoom level. "Small" means that the features are usually not bigger than 1 tile in the single zoomlevel that they are served on. The features in the dataset are validated beforehand (self intersecting, etc). Clipping, simplifying and cleaning/validating are not necessary in our case, and disabling them increases performance. Especially cleaning takes a lot of processing time, which takes the response time above a workable threshold for this dataset. (Especially when clipping is disabled, but that's a known issue, I believe.)
tl;dr: an option to disable cleaning/making valid