The other thing to note is the system we use for the high resource stuff -does- work, but it also doesn't suspend instantly when someone exceeds the limits - if it did, then pretty much everyone on free hosting would be suspended, even if they used more than the limit on CPU for 0.01 seconds. It does allow some tolerance cause often an app uses a lot of cpu but finishes running in under a tenth of a second.
The -real- problem is after suspending someone who was abusing resources, the server's got a backlog of requests to catch up on from when the user was abusing the service. It takes a bit for it to catch up, especially since there's now more requests coming in from people mashing refresh
Not saying it was actually abuse that caused this slowdown, just pointing out if it was why it doesn't recover instantly.