PCI Resource Error (MBP 2016 15”, 19.10)
just as a disclaimer, I have not had the time to systematically examine this, so take the following with a grain of salt:
The PCI bridges I am removing are the 8x (0000:00:01.1) and 4x (0000:00:01.2) bridges on my system. These bridges seem to be the ones managing the Thunderbolt Ports on either side (8x left, 4x right I believe). There is also a 16x PCI bridge (0000:00:01.0), however removing that one will most likely crash your Mac as it is responsible for most of the internal connections.
From taking a rudimentary look at the logs, this is essentially what happens: Removing either bridge also removes their Thunderbolt Bridges (in my system 07.00.xx I think). Upon rescanning the system then reports freeing up resources from these ports, which I believe then allows the card to be recognised properly. The inspiration for this came from this post on GitHub, although my specific order of commands just came to be through testing. So you should play with it, removing only one bridge, rescanning, checking what dmesg reports, etc.
In regards to your issue with pci=realloc, that is fairly interesting, I never had that issue. I would check if the Core itself is recognised (either in the System settings under Devices or with the command boltctl). If not, try the process above without the parameter, I am not sure if it is essential in the first place.
Lastly, I have not attempted to use eGPU-switcher or gswitch with this new setup, as my main concern for now was CUDA support and that works flawlessly and does not require the GPU to be a display output. However, if you get the card recognised by the driver, getting it to work with eGPU-switcher or gswitch should not be a major issue, as the process should be very similar to the one followed by others.