Honestly, splitting sources often backfires more than it helps.
Meta just sees two numbers both getting hammered by cold inbound from people who haven't even saved the contact. Different lead sources, sure, but the traffic pattern looks identical to the system - high volume, zero chat history, no warm signal whatsoever.
WhatsApp has no way of knowing this is legitimate business activity. The API can handle it because you're sending templated messages through an approved channel where Meta already expects volume. But the regular app? It's basically screaming "spam" at the algorithm.
Also worth double-checking if either number was used for outbound blasts at any point. One cold blast can permanently flag an account, even if it happened months ago.
Were both numbers set up around the same time?