Instrumented Mk4 TRNG read count

Shiny reports instrumenting Mk4 firmware at the point where it reads the STM32 hardware RNG and then requesting the same 32 bytes seed generation uses: on 5.6.0 exactly eight hardware reads were observed, which the poster reads as direct proof that the seed request reached the true RNG, and on affected firmware the same test is said to show zero hardware reads with the bytes coming from MicroPython's software PRNG instead. The post frames this as a real-device path test rather than statistical inspection of whether output looks random, and asks why such a test was not already in COLDCARD's suite and run routinely on hardware. It quotes the vendor's reply to another account insisting there was no weak entropy fallback and that precision matters, so it is a direct methodological answer to COLDCARD on what a test would actually settle. No instrumentation code, device identifier or log is published with it, and the affected-firmware half is stated as what the test would have shown rather than a run that was performed. The result and the inference are the poster's own and are not reproduced here.