Comment 100 for bug 254668

@Martin: At least the log you added does not seem to be from a run with noapictimer. If you could do and post the dmesg of two boots. Both with "apic=debug debug" (acpi=debug is ignored with the default kernel). For one add "noapictimer" for the other remove it. Then post both dmesg logs (if possible not compressed, just the plain files. makes it simpler to peek on them).
As for prevention: you could also play with "nohz=off" and "highres=off" at the cost of higher power usage. This worked for some. Or try the combination from the upstream bug "noapic lapic acpi=on" to see if that helps. Of course this still is no solution but collecting hints.