I found that running rmmod acpi_cpufreq modprobe acpi_cpufreq after resuming from suspend resolves both this bug and #505180 . Perhaps suspend could be configured to automatically reload acpi_cpufreq on resume.