A planned hold adds flexibility to the schedule. The hold at T-4 was planned to be 10 minutes long if everything went without a hitch. If a problem had cropped up (e.g. weather suddenly deteriorates and goes outside the launch limits), the hold would have been longer. If there were delays earlier in the countdown, the T-4 hold might have been shorter.
The sum of all planned holds plus the countdown time is calculated to get to the start of the launch window. The planned length of the final hold varies (25 minutes at this Atlas launch in 2005, Shuttle launch transcript showing a variable-length hold at T-9 minutes).
The hold at T-4 has these purposes:
- NAM and NLM final launch polls -- go to continue countdown
- Spacecraft transfers to internal power
(see the Atlas V launch overview). I watched the MAVEN launch coverage, and I don't remember the final launch poll taking up all of the 10 minutes allocated. So they finish the poll, then wait until they can restart the count so that T-0 coincides with the start of the launch window.
When the clock is restarted at T-4, these events happen:
- Automatic computer sequencer takes control for all critical events through liftoff
- Atlas first stage LO2 replenishment is secured, allowing the tank to be pressurized for flight
So when the clock ticks past -4:00, you're committed to launch within a reasonable time, before your LO2 tank boils off too much O2.
This is speculation: the reason for the planned hold at T-4 might be psychological. Nobody likes to make big decisions under a looming countdown.