First, I am very glad the you have alerted people to the pints|gallons relationship of bits|bytes. Your numbers show you've had some sad experiences with 10Gbps connectivity, and that you've experienced the architectural limitations within Thunderbolt.
A 10Gbps connection has a maximum ceiling of 1250MByte/s, and in real-world usage will cluster around ~950MByte/s on an SSD, peaking near ~1100MByte/s even across an 10GbE network connection. If you're seeing 350MByte/s throughput, there's a strong likelihood that something got downshifted to 5Gbps.
The Thunderbolt architecture always had PCIe data and video DisplayPort reserved bandwidth. The PCIe portion is/was around 22Gbps, so that explained the ~2800MByte/s SSD throughput typically experienced by users. The TB4 and USB4 specs have moved the reservations within the 40Gbps envelope, and now numbers exceeding 3500MByte/s are possible; Corsair's USB4 product advertised 4000MByte/s ... which is 32Gbps.
Thunderbolt 5 products have started to arrive, advertising 6700MByte/s throughput (~54Gbps) so things are great in the drag racing game of SSD benchmarks.
You've also apparently had some sad experiences with HDD's. I *never* see less than 120MByte/s on even my oldest 3TB Hitachi drives. Average for those is 150MByte/s, and the newer 8TB Seagate drives run nearer to 200MByte/s on a single-drive basis.
The best takeaway from your observations is that everyone should build to their needs, and avoid using benchmarks as the single decision criterion. We're now in a world of products that allow us to optimize for the things that affect our interactive workflow and delivery constraints. For example, while it's interesting to have a super-fast Time Machine backup capability, I don't think the speed investment improves workflow or the economics of running a computing infrastructure. Even comparing the benefits of a 4GB/s SSD vs a 2.8GB/s SSD should give us pause to consider what would be better by displacing the older product. If it's a new-build, then by all means, get the new/faster one if the cost-per-TB is close enough, but for storage, capacity and reliability should be always be paramount.
Sorry for the soapbox; the coffee was extra good this morning.
