QoS on MPLS networks (L3VPNs)

Hi community.

I’ve been lately trying to study a lot the SP part of our Jobs because this is the thing that I enjoy the most, quite complex and huge I have to say. Currently studying a little bit about QoS, and then I came to make myself this question:

Say there’s ISP-A, this ISP offers L3VPN services over MPLS to its customers, this ISP provides 3 levels of QoS to all its L3VPN customers:

A. - Gold (low latency) B. - Silver (regular latency, low discard) C. - Bronze (best effort)

When a customer purchase this service he gets per site “a guarantee” of bandwidth, say it’s 10 Mbps and also gets to tag its traffic to be treated like either of the 3 tiers aforementioned. My question is, what does this mean in real world?

I think (please correct me if I am wrong) that:

A. - QoS for this customer shouldn’t be needed in the core links, because if you’re providing me a bandwidth guarantee you must ensure that I have all this bandwidth 100% of the time available

B. - In case QoS is needed/used, then how is it done? I mean, I highly doubt providers go ahead and program their core links hop by hop with each customer needs, and I doubt this is even possible due to the limited amount of bits the EXP section of MPLS has. And also, wouldn’t this mean they’re just not providing the necessary bandwidth as contracted by its customer?

So, what I am thinking is that the QoS is basically only applied to the traffic as it exits the MPLS cloud (as it transitions from PE to CE) due to the nature of L3VPN (any to any) this would be needed to avoid priority traffic being dropped when a lot of sites decide they need to send traffic at the same time to a single site. Am I correct here? Is this the typical approach?

I have to admit, I do have a hard time trying to grasp everything QoS related in the SP world, especially with the considerations regarding multi-tenancy

Thank you in advance!