This is probably due to a return route difference. Because we're only showing the route in one direction (not the route the data goes through on the return route), it may be that some hops / data types have different participating routes. The 8 to 9 link *might* be the problem in this case, but it may also be that data coming back from hop 9 is following a different route than the data coming back from hop 8 - and that this is actually the issue. This scenario starts to get pretty challenging to troubleshoot, though, unless you have control over a destination that is showing these problems (because you can then trace / pingplot from the opposite direction and see which route *is* participating, and where latency is being added in the return route).
You might start this out by talking to your ISP. They might be using multiple providers / peers, and they may know that one of them is getting saturated.