We had this same issue. We have a large table with much indices and some day all goes slow.
Reducing index quantity helped a lot.
We was using 90 pages of cache in Classic. We increased this number to 150, shutdown all conections and thats it! Problem solved.
Latter we increased this number to 2048 as Superserver default. Our big client has 140 active simultaneous connections and no problem. Of course we calculate the necessarie ram.