I have now, twice, with a distance of a decade between the experiences, seen a VPS provider's simplicity and ability to easily scale stop mattering at high-ish but
not extremely high levels of traffic—in both cases, because their network couldn't keep up with traffic, such that we were finding ourselves network-bound even on pretty weak VMs, everywhere. Apparently poor inter-connects or peering agreements or something (not sure) leading to consistently terrible performance for people in some regions, heavier public traffic hitting a wall long before a smallish VM was starting to sweat, that kind of thing.
First time it was Rackspace, second time it was DO. First time we fixed it by leasing bare metal somewhere (I forget where) and running our own libvirt/KVM VMs, second time we fixed it by switching to AWS. In both cases, that fixed all the network issues.
AWS has a separate problem that they have these vague classifications for network quality and it's hard (at least, it was for us) to feel certain the network was the problem, but at least it's not that hard to spin up a VM with a better-rated network to trial-and-error one's way to the right setting.