Hello Sims
In the example you describe, I am assuming that you have a layer 3 switch for example, where you set up a QoS policy on the SVIs of each individual VLAN being used internally on the network. (You can also use a router with subinterfaces, each on one of the mentioned VLANs). Also, I am assuming that you are talking specifically about shaping or limiting the traffic on each VLAN as indicated in your text.
So if you have a shaping or rate limitig policy on each of those VLANs, then the policy will be applied only to traffic that goes through those interfaces. This information is not conveyed in any way to the ISP router, so the same traffic that traverses the ISP router and enters your L3 switch will not be policed at the ISP. This means that if additional traffic comes through the ISP greater than the shaping limits, then the L3 switch must take care of this extra traffic using its interface queueing mechanism, that is, memory allocated for packet buffers at each interface. If the queues are saturated, then packets will be dropped.
I hope this has been helpful!
Laz