Deferring indexes/constraints to be created in parallel with CDC

[agis]

Hello! In trying to speed up the process of cloning a 10TB instance, we thought of deferring index/constraint creation to happen in parallel with the CDC process.

I'd like to run this by the community to see if there are any caveats or issues that you can imagine when doing this. Does this sound like a reasonable optimization? Does it come with any data consistency or other caveats?

Thanks in advance!

P.S.: One thing that comes to mind, is that CDC might be impacted given that it may probably not have yet the PK indexes available to locate the rows on the replica.
P.S.: An implementation can be seen here.