I’m afraid we’ve had a miscommunication here about this issue, it turns out that what you’re seeing is actually expected behavior. Sorry for the confusion, I’ve been learning about this today too!
I’ll explain in detail below, but the correct answer is that you cannot change a send level from -inf dB using a blending snapshot. Using an overriding snapshot is the best option if that’s suitable, if you require some kind of blending behavior then you would need to arrange it so the base send level is not -inf dB.
So it turns out that the send level is modelled the same as a volume, which makes sense because a send is basically a fader which routes to another bus. And the way send level works with a blending snapshot is the value in the snapshot is added to the send level when the snapshot is active. This allows multiple blending snapshots to act accumulatively.
E.g. if your send level started at 0dB and you had two blending snapshots which ducked the send level by -3dB, then activating one of the snapshots would duck the send level to -3dB and activating both would duck it to -6dB. Boosting behaves the same way so if you added a third snapshot with a send level of +2dB then activating all of the snapshots would result in the send level being -4dB (0dB base level, -6dB from the first two snapshots, +2dB from the third snapshot).
There is an additional important wrinkle when any of the levels is -inf dB the result is always -inf dB. This includes the base level being -inf dB, which brings us back to your situation. We can use as many blending snapshots as we like to add as much as we want to the send level but if we start at -inf dB we’re going to stay at -inf dB.
Overriding snapshots worked when tested because they directly set the send level to whatever value the snapshot has.