Thunderbolt3 devices
--  OWC Express 4M2
--  UNIBOS Portable SSD x4

finally provided clean pnpinstal of Intel 769v66 drivers
then ... update to 789
then ... update to 987 (with minor manual fix of Toast device missing + appInstall)
(from now, Thunderbolt downstream/upstream description is missing,
 there are ONLY generic "PCI-to-PCI-bridges" !!!)

AND: it seems that earlier thunderbolt host controller drivers v769 and even v789
behave differently to difeerent SSDs in regards to "disk policies", as
most of them was connected for "quick removal" by default so there was NO CACHING and
very slow 4K performance - only KINGSTON KC2500 seemed originally "as the best", although
may be this one was ALWAYS connected with caching enabled by default !!!! (so dangerous)
(its not even clear if this was visible in device manager, I originally tested all
SSDs expecting that ALL are mounted for "quick removal", not knowing the 4k speed
differences in this case - once it was clear, I started to check also behavior of SSDs
during enabling/disabling cache - here both WD SSDs REQUIRE restarts but inconsistently
in fact, WD Black SN-750 needs restart to enable, while WD Blue SN-550 needs restart
to disable caching (both checkboxes always, win+device write)
 
since thunderbolt host v987, NO SSD is connected this dangerous way, so even KC2500
seems to be "slow", BUT of course, when windows and device caching is enabled in
device manager (see restart requirements above), then EVERYSTHING IS FAST !!!
as expected, so also 4K acces !!! and of course on ALL SSDs tried .... 
everything is similar, good.

ALSO, v987 has NO BSOD issues it seems, and also allows quite safe EJECTING of SSDs
followed by manual device manager root (OWC) (downstream) PCI bridge DISABLE/ENABLE
cycle to detect new connected SSD (unfortunatelly, simply inserting TB3 connector
doesnt prompt PNP device manager detection of new device (with BIOS PCI pwrmngmt
disabled, at least), BUT DISABLE and REENABLE of root (downstream) PCI bridge
does the job and new disk is properly detected, without ANY issues. It seems that
behavior of THIS latest v987 thunderbolt driver is finally PERFECT with currentlly
safe working generic MS "nvmestor" NVM Express driver.

Will be interesting to test latest Update KB with new versopn of vmestor 19041.488, 
how this will finally behave, but FOR NOW ITS PERFECT FINALLY !!!

---------------
So, after installing latest KB4571756 (with stornvme 488), everything WORKS as with
older driver, it seems till now, EXCEPT in combination with Intel 660p fw004C !!!
(not sure about caching enable/disable restart requirements, it was NOT required it
seems on portable/plugable UNIBOS device, will try it again, probably by other
uninstal of latest KB)

Another interesting thing happened in devicemanager after enabling BIOS PCI power management
settings "PCIe ASPM Support" and "Native ACPI OS PCIe Support", where BIOS hint notes,
that it may be unsupported on all devices but relates to power and performance.

When both this BIOS setting was disabled, device manager showed ALL "PCI-to-PCI Bridges"
inside thunderbolt deices this generic way. After enabling BOTH of ettings, device manager
started to show "PCI Express Upstream Switch Port" and "PCI Express Downstream Switch Port"
instead.

-------------
So, finally, enabling ONLY NATIVE OS ACPI does the job and allows proper PnP attachement
after restart (no powwer cycle required, everything is stable and consistent on each disk)
(Except of cource Intel 660p AND sometimes ALSO DVDV BSOD still happened on OTHER disk,
occured on KC2500 at least, but very randomly, probably while extreme hurry to PnP or manual
enable-disable of downstream in device manager)