What I think I discovered is that this is on Google. Since Pulumi uses the Google API to delete the node pool, what Google does is to cordon off every node in the pool, and then start to delete each of them. I believe this ignores the PDBs.
It would be great if there was a way in pulumi to have it handle it via a rolling delete. Especially given that due to being pulumi, you can’t have 2 resources with the same name. So you can’t have a new node pools ready ahead of time.