I'm afraid that there isn't enough information in that thread to help me understand what the problem is and how to reproduce it.
Generally speaking, the mechanism that computes the "silence status" of Photon compares input sample values against a silence threshold provided by the RE SDK. The silence threshold value is "hard-coded" into the compiled RE byte code that I send to RS for acceptance testing, so the "logic" of the silence detection does not change with the Reason version.
If Photon's silence detection does behave differently in R10 compared to R12, then the most likely reason is that the device that generates audio input for Photon has changed its behaviour. Or the Reason hosting may have changed its behaviour in some way. Another reason could be that the RE build server that compiles the RE byte code into Mac and Windows assembler uses a different compiler version and/or compiler flags when it creates the R12 hi-res version of the RE.