Right now, you can pick only instances with SSD size >= from the current one.
The issue is that sometimes you need more CPU, sometimes you need more RAM. If you ever happened to use Space optimized droplets there is no way you can scale back for reasonable money. You may need less CPU at one point in time. There are lots of scenarios when you really need it.
My proposal is to allow downscaling of instances automatically.
Right now you can achieve it by creating a new smaller DB and migrating all data from the old one.
I think it can be fairly easily implemented in a managed manner (just create a new, smaller instance, migrate data, destroy old instance)
You may put a condition that the new instance must have enough space to accommodate current disc usage +30% for safety reasons.