Thanks for the report!
Yes, it really seems diode_connections.lvs is the only place where nwell is used for the connection to the devices.
Everywhere else, i.e. moscap_extraction.lvs/mos_extraction.lvs/mos_sab_extraction.lvs/res_extraction.lvs, nwell_con is used.
If you take a look at general_derivations.lvs you can see that nwell_con is defined as:
nwell_con = nwell.not(res_mk)
I think this is so that nwell resistors are correctly extracted. They are extracted as:
# NWELL
logger.info('Extracting NWELL device')
extract_devices(resistor_with_bulk('nwell', 1000, BResistor), { 'R' => nwell_res, 'C' => nwell_con, 'W' => sub })
You can find a description about the extractor here: https://www.klayout.de/doc/manual/lvs_device_extractors.html
Taking a look at the connectivity setup in general_connections.lvs, only nwell_con is connected to ntap:
# Inter-layer
connect(lvpwell_con, ptap)
connect(lvpwell_con, ptap_dw)
connect(dnwell, ntap_dw)
connect(nwell_con, ntap)
connect(ptap, contact)
...
That's the reason why you get anonymous pins with just nwell.
Long story short, I agree that nwell_con should be used :)
You might want to use the LVS setup in this directory anyway, as @Clyde Laforge has made some important improvements around substrate cuts recently.
We are working on creating a new open_pdks PDK, which will allow us to switch from the current wafer.space PDK fork. However, this probably won't happen until next week, as I'll be at Hackaday Europe this weekend :)
Awesome, thanks for taking a look @Leo Moser (mole99)! No credit needed, probably easier/faster for you to make the fix than for me to cobble together a PR
Will take a look at that LVS setup for my own experiments! and goodluck/have fun at hackaday conf
Discovered thanks to a report by Zachary.
diodenw2ps* used the wrong layer for the bulk connection. general_connections.lvs uses nwell_con to connect to ntap.
Will take a look at that LVS setup for my own experiments! and goodluck/have fun at hackaday conf 
1
1