This error is rare, I had seen it few times over several months of testing. I happened twice in the last 2 days, both times at night.
It happened with OS 35.3.1 and earlier.
I found that in those few cases that I managed to obtain syslog from there is a correlation between message
systemd-timesyncd: Initial synchronization to time server 22.214.171.124:123 (1.pool.ntp.org). which is followed by kernel: [ 46.450236] nvgpu: 17000000.ga10b ga10b_pbdma_handle_intr_0_legacy:437 [ERR] semaphore acquire timeout! kernel: [ 46.460468] __ga10b__ Channel Status - chip ga10b kernel: [ 46.460470] __ga10b__ --------------------------- kernel: [ 46.465329] __ga10b__ 420-ga10b, TSG: 26, pid 2879, refs: 2, deterministic: no, domain name: (default) kernel: [ 46.470176] __ga10b__ channel status: in use idle not busy after that __ga10b__ errors are printed non-stop and Orin requires reboot. However, message "Initial synchronization to time server" appears in good boots as well, but in bad cases it always precedes the __ga10b__ error.
What can it be? Is it possible that “semaphore acquire timeout” somehow causes by system time change?
I see that other people reported this error after system resume, but I never performed any suspend or resume, just reboot using “reboot” command.