Unfortunately I don't believe the addition of /sys/power/wakeup_count is sufficient.
While it does allow us to know if any event has occurred before we try to suspend, it doesn't tell us the event has been consumed by userland.
The easiest example for me is my nightly backup, which is triggered by a RTC backed alarmtimer is scheduled for midnight. Then I walk away from my computer at exactly 11:45, when the system is set to suspend after 15 minutes of no X input.
The RTC interrupt fires first, upping the wakeup count and waking up my backup application. Then before my backup application can run, the pm-daemon starts, reads the wakeup count, echo's it back in, and then echo's "mem" to /sys/power/state. In that case, the system would suspend and by backup wouldn't start until I returned to my computer the next day.