[Bug?] CUDA kernel information lost after CUDA profiling data flush

I am profiling a Jetson AGX Xaxvier with Nsys 2020.5.3
I have enabled the CUDA trace collection, and found a weird phenomenon.

In the image, one can see that after a particular profiler overhead, the kernel names are replaced with a generic description “cudaLaunchKernel”. It seems as if the kernel name information is lost.
One can also see that the “CUDA HW (Xavier)” timeline has been cut off at that particular profiler overhead.

Is this a bug or am I doing anything wrong? The attached file is my report.
Report 22.qdrep (8.4 MB)