The general pattern looks good, although there will be some details to talk through and some docs about how to manage partitions, maybe some builtin changes to do it automatically etc. The migration generator is going to have to have some kind of handling for changing the partition strategy as well.

I noticed some unfinished bits as well, i.e for creating range partitions. We should definitely avoid the string interpolation in the SQL there as well.

I made some updates before reading your comments.
I will try to make a new draft considering your comments :)

Also found some problems when updating the primary key on multitenancy resource.
Not sure it this is how you would prefer to fix it.

Could we separate out that other fix that you mentioned into its own PR so that I can review in isolation?