Possible limiting factor during upgrade of Fedora VM - not disk or cpu or network?

I'm upgrading a VM from Fedora 29 to Fedora 30. I triggered the update from inside GNOME Software. It downloaded the upgrade, and asked to reboot. It started installing... atop shows the qemu process using about 15% of one CPU core. And my one hard drive at about 15% utilization. What other resource(s) could the VM possibly be waiting for, the other ~ 70% of the time? Is it just [sleepy](https://en.wikipedia.org/wiki/Sleep_(system_call)) ?? Is there some other way to explain or find this out? The upgrade showed the normal progress indicator, and completed without errors. The VM is in libvirt (virt-manager), using the default KVM backend on x86-64. I configured the VM disk using cache mode "none". There's a 2x2 matrix of caching options; "none" means O_DIRECT, but not O_SYNC. (My intent was to avoid [filling the host kernel's writeback cache](https://unix.stackexchange.com/questions/474119/system-lags-when-doing-large-r-w-operations-on-external-disks/474171) ; at least it seems to have left the host system so much more responsive than the last time I tried this). "IO Mode" is set to "Hypervisor default". The virtual disk controller is using VirtIO. libvirt is from Fedora 29; it is version 4.7.0 and qemu is version 3.0.0.