I can answer this, since I fully understand how this sploit works (and wrote it independently myself).
They did not pwn TSEC at all. What the sploit does is fool TSEC into thinking it's running with sole full control of the system (it thinks the CPUs / DMA are halted). Only then will it continue to generate the keys and decrypt the package1. To fool the TSEC, it needs to read the same constant values from MMIO memory space. This can be done by remapping the address space into DRAM by using SMMU translation. If a single bit from this space is different (like say the BPMP CPU is still running, the IO space would reflect that) then the TSEC would detect it. But by mimicing the same values in DRAM, the TSEC is fooled and continues decrypting package1.
To actually pwn TSEC you need to get code execution in the authenticated mode of the TSEC, which can then be used to reveal TSEC secrets. To date, nobody (and I'm pretty sure not even reswitched or switchbrew) has managed to do this.