Need to call prepareToPlay twice to get audio graph to handle latency ops

I have a graph which incorporate multiple nodes, some have oversampling.

The issue is that prepareToPlay must be called on the graph twice in orders for the rendering ops to be constructed correctly for the graph.

This is because the graph builds the ops – and then calls prepareToPlay on the nodes, so their latency hasn’t yet been set at this point.

Calling this again uses their current state so the latency is set and this works correctly.

Curious if there are other methods for this? Previously I used customized version of the JUCE graph, but I’m tryin got get off that.