I think there's some strong hints in apple's implementation that they were under external pressure.
If you read the documentation on how it's implemented, there's some fairly advanced crypto - private set intersection, threshold secret sharing - that only makes sense if Apple took the line "we have to do this, but we're willing to do it in a really expensive way so that we ourselves have as little access as possible". They went to the effort of running the NeuralHash on the client device, as far as I understand.
The standard implementation on other cloud providers is that the provider has access to your data on the server if they really need it - much cheaper, and makes it technically possible in future to easily change the T&C to "we may use this for market research and to improve our products". I view Apple's client-side implementation as drawing a very big line in the sand saying we will not do that and we are willing to put our money where our mouth is by writing this crypto protocol.
But that certainly raises the question, why would Apple do anything like that at all, unless there's some (current or anticipated) external threat that forced them to?
As to whether we'll ever discover it - I wouldn't have predicted the Snowden revelations, maybe this will come out the next time something like that happens. If it does, I also predict we'll be shocked to discover just how big the market and distribution network for child abuse online is, and how many people are involved.