Many thanks to @chrisyuan for pulling out some details from various bugzilla’s around the web. I confirm this works for the WDS512G1X0C as well.
The 5500us is based on Samsung controllers (PM951, PM961, evo 960 etc) having P3 power states that report their entry and exit latency for P4 (the lowest Autonomous Power State Transition or APST) at ~1,500/6,000us meaning that 5,500us will enable P3. On at least my WDS512G1X0C (slightly different than the newer 3d NAND 2018 model this thread references) with firmware version B35500WD the entry/exit latency for the 3 power state is: 51,000/10,000us (yes, those numbers seem awfully fishy to me but they match with 4 at enlat:1,000,000 exlat:100,000). This is from
sudo nvme id-ctrl /dev/nvmeX where X is the nvme device in question (often 0). Certainly 5,500 and limiting the drive to P0, P1, P2 works but P3 might be achievable based on many other people’s experience with Samsung controllers.
The real issue COULD be APST or PCIe Active State Power Management (ASPM) or an interaction between the two. Others have had success with firmware updates from their manufacturers however as we’re in Linux not Windows there’s no way to tell if there is a firmware update for the WD Black NVMe drives.