Battery-operated devices don't poll immediately. If a battery-operated device was constantly on the ZWave network listening, it would pretty quickly kill its battery. Not as fast as if it was constantly transmitting, but even running with the receiver on is an issue. What most devices do is sleep the CPU and turn off the radio until either an event happens that they care about (like the door they are monitoring being opened or closed), or a timer expires. When an event happens, they do what you expect them to do, like any other device: send a message and wait for an acknowledgement. Then they go back to sleep. When the timer expires, they "wake up" and send a packet of information (or several), often including state and battery level information, maybe temperature, humidity, etc. if they have that, and then go back to sleep. It's a "just wanted to let you know I'm still here" message.
Because these devices are not on line constantly, you can't poll them. The chances of you (or Luup) hitting the device at the moment it is awake for whatever reason are about as good as getting hit by lightning on a sunny day. At some point, the device wakes up and tells Luup what's going on, so this serves the purpose of polling, it's just driven by the device rather than by Luup.
Configuration is actually no different. When Luup starts, the device is marked unconfigured/pending, and it's not until the device first wakes up and communicates with the hub that Luup can then see the device needs to be (re)configured and does that. Then the rest of the dialog happens. So unlike your mains-powered devices, your battery-operated devices may not even be configured for hours after a Luup startup. This is also why many devices have a "wake up" mode you can start, often by pushing a button briefly or several times, that keeps the device awakened and talking for a short period of time. This feature is particularly helpful right after inclusion when you are first configuring the device, and you should always look to see if the device has this feature and know how to access it.
The "wake-up" interval is settable in the UI for many devices. Sometimes, the device default wake-up is huge, like 86400 which is 24 hours. That means the device will only talk to the hub once every 24 hours unless another event forces it to. If you try to change the wake-up interval, you may have to wait for the current wake-up interval to expire so that the device will start talking and the hub can tell it to shorten the interval (another reason the forced wake-up mode is handy -- removes that delay).
I suspect what you are seeing is that Vera may wait for a couple of missed wake-up intervals before it decides the device is unavailable. Setting a shorter wake-up on the device may reduce that time, but keep in mind, more frequent wake-ups mean more battery use, means less battery life.
P.S. It should follow, then, that setting Vera's polling on a battery device is pretty much useless, and empirically we've proven this. You can usually turn polling off for battery-operated devices with no ill effect. There are, of course, some exceptions, so experimentation and observation are always advised. But just remember, polling is initiated by the hub, and wake-up is initiated by the device. They are not the same.