This really depends on the size of your blobs. Having a small jsonb payload will not be a problem, storing 500MB blobs in each row is probably not ideal if you frequently update the rows and they have to be re-written.

Doesn't Postgres actually use it's TOAST system (basically pointers to blobs on disk) for any large value anyway?

Ominous... Care to elaborate?

Why?

more deets?

I want to use Postgres for JSON, I know it has specific functionality for that.

But still, what do you mean by that and does it apply to JSON why or why not?