Yep, but you wrote this:
Which left me with the idea that direct disk access is still possible (even after running that command) .
Surely you do realize the difference between formulating hypotheses as part of an ongoing investigation where you have yet to figure exactly how to address a behaviour you're observing, vs. eliminating said hypotheses as you are zeroing on the most likely explanation.
I played a lot with win32diskimager and Rufus initially, and saw some conflicting behaviour there, which I tried to make some sense of. So I jolted down some notes of what might be happening and what I might try next (which I had yet to test). But as one progresses along, and as is expected from the scientific method, some of the earlier statements may find themselves having to be refined, or simply thrown out. So if you hold too much credit to a "let me jolt down some notes before I go to bed" note, I'm going to be very unhappy about having to explain to people that what they see on an issue tracker is obviously a very fluctuating, raw and ongoing deduction process...
In other words, whenever you see something that seems contradictory as part of an investigation, it's obviously the concluding/most recent statement that needs to be taken into account.
As you may also expect, disconnecting and reconnecting the device on the same platform is one of the first thing I tried, but that was to no avail after it had gone through an IOCTL_DISK_CREATE_DISK cycle. Maybe, as you suspect, this is due to Windows keeping a record of the device somewhere, or maybe it's something else. At any rate, I'd encourage people who are interested to play with the steps I provided (that's why I provide them), to try to shed some more light on the issue.