Odd? Major Database engine (Oracle for example) have "CREATE SEQUENCE" for just this purpose
The idea is that each of those dozen tables has a specific purpose (they each describe a different type of data)...
and some of those tables hold data where each record relates to a different table
For example Table D.... one record might relate to a record in Table A, while another relates to a record in table B
there for the "key" cannot be an AutoIncrement PK in A or B since that would overlap.... so the PK values in A and B must be unique across both tables, so "Where A.key = D.key" never conflicts with "Where B.key = D.key"
An example.... I used to work in Healthcare.... and we had (for sake of discussion) one table that was HOSPITALS, another that was DOCTORS and yet another that was CLINICS... each of these had a unique record structure..... but they all had some things in common... so those were in single tables regardless of provider type. That is a super simplified example (in reality it was much more involved). My project consists of hundreds of records... not millions