Hello, Manu!
Thanks a lot for the review, especially for the rebase and the
in-flight INSERT case.
The logical replication part has had its own thread for a while [1], so
I've replied there and posted v20 there as well.
In short, the behavior change is real and is now described in the commit
message. I think master only happens to win a race in that case rather
than providing a guarantee, so v20 leaves the wait out. More details
are in [1].
[1]: https://postgr.es/m/CADzfLwXZVmbo11tFS_G2i+6TfFVwHU4VUUSeoqb+8UQfuoJs8A@mail.gmail.com
Best regards,
Mikhail
