A soft bounce says not now. The mailbox is full, the receiving server is temporarily unavailable, the message was too large, or you are being deferred because the receiver wants you to slow down.
The correct response is to retry with backoff — and then, after enough repeated failures, to give up and suppress.
Deferrals are the interesting case
The most consequential soft bounce is a 4xx deferral issued because a mailbox provider is throttling you. It is not a problem with the address at all; it is a statement about your sending.
Google’s own guidance on this is unusually specific: wait 15 minutes, test with a single message, stay below the volume that triggered the deferral for 24 hours, and only then increase again by 25–100%. Pushing through a deferral by retrying harder is precisely the wrong move and reliably makes it worse.
Where the threshold sits
There is no universal rule for how many soft bounces should become a suppression. A reasonable default is to suppress after the same address soft-bounces on several consecutive sends over a few weeks — long enough to ride out a holiday mailbox filling up, short enough that a dead-but-not-yet-rejecting address does not linger for months.
Across several brands
A deferral is issued against your sending IP and domain, not against a brand. If several brands share a sending domain, one brand’s oversized send produces deferrals that delay every other brand’s mail through the same path.
Separating sending domains per brand contains that. So does setting per-connection rate limits and concurrency deliberately rather than letting each brand send as fast as the provider will accept.