No neutrality rules have been broken because the ISP isn't shaping or filtering.
But the ISP would be failing to deliver the promised bandwidth. If the ISP would always deliver the bandwidth they had promised to the customer, there wouldn't be an issue. Unfortunately the ISPs will always pull the disclaimer about not guaranteeing, that the server you are accessing has spare capacity. Though this disclaimer makes sense, it isn't necessarily true in all cases, where the ISP would apply it.
If A want to send packets to B, and if A is not using all of the upstream that A has purchased from their ISP, and if B is not using all of the upstream that B has purchased from their ISP, then packets from A to B must get through with no packet loss caused by congestion. If there happen to frequently be congestion between the two ISPs preventing packets from being delivered even though neither endpoint is using all their capacity, then the ISPs are simply not delivering, the capacity they sold. And the ISPs should be required to make arrangements to upgrade capacity to match what they sold.
I only consider application of the disclaimer about the capacity of the other endpoint of the communication to be valid, if the other endpoint is actually using all of their purchased capacity. Simultaneously using that disclaimer against both endpoints of a communication smells like fraud.
There are other aspects to communication than the bandwidth. Packet loss and latency are just as important, but they are rarely advertised. The latency between two endpoints must never exceed the sum of the latency advertised to each endpoint and the latency inherent to the physical distance between the endpoints. AFAIK you can expect about 1ms/100km of roundtrip latency with the speed of light in optical fibers. In other words, if A has bought a connection with an advertised 5ms roundtrip latency, and B has bought a connection with an advertised 10ms roundtrip latency, and if the distance between A and B is 10000km, then the overall roundtrip latency must be not more than 5+100+10 milliseconds. Exceeding a roundtrip latency computed this way is not acceptable, not even due to buffering. Additional buffering would be acceptable if the sender explicitly picked a ToS specifying a desire for additional buffering, but it is not acceptable on the default ToS.
Finally packet loss should not exceed the sum of packet loss specified on the connection for each endpoint, unless either or both endpoint is exceeding their purchased capacity.