Hmmm... that suggests that the problem is probably outside of Roastmaster. Not 100% of course, but the same code runs to start/stop probe functioning, regardless of whether it's being run at the behest of the roasting console for a roast or just that simple probe details view.
Since you mentioned restarting your device, one thing to note... when you tap the home button to quit the app, it doesn't actually quit - it just transitions to a background "idle" state. If a probe is sampling data for a roast in progress (not just displaying its temperature before or after the roast), Roastmaster will continue to actually run in the background, otherwise it just goes dormant. You can physically terminate Roastmaster's execution entirely by tapping the home button, then double-tapping it again to see the background apps, then swiping up on Roastmaster's icon/screenshot. That should have the same net effect on Roastmaster that a complete device reboot would - it would clear out all Roastmaster memory, and the next launch would be fresh. There is nothing I can think of that a complete restart of iOS would gain for Roastmaster's functioning, besides a little time for the probe to heat or cool and resolve the electrical shortage, IF that is in fact the issue.
But before we jump to that conclusion, another thought: the newer iCelsius SDK versions will auto-launch the iCelsius brand app when you connect the probe. This could be due to functionality that became available in iOS 7.x - not sure, but that might explain why you've never seen this before if you've recently upgraded. When their app auto-launches, most folks will just tap the home button to (assumedly) quit it, then launch Roastmaster and perform a roast. Here's the kicker - the iCelsius SDK (sometimes) does not like sharing the probe between two apps. If the iCelsius app is open on the device (even in the background) it can often commandeer the probe, and Roastmaster will be unable to connect to it until it relinquishes its lock on the probe. Since a restart seems to help you, make sure their app is not open in the background. If so, remove it from the background tasks in the manner described above. Note: Once their app "commandeers" a probe, Roastmaster might need to be restarted in the same manner to get the SDK to realize the lock on the probe has been released.
If that doesn't resolve it, it might be helpful to conduct a few dummy roasts with the probe at room temperature. Just let them run the normal length of one of your roasts, then conclude them in Roastmaster and see if you experience the same connectivity problem. That will help rule out a short in the probe.
Please keep me posted. I don't recall seeing the "commandeering" issue before the current version of their SDK, and a lot of folks new to data logging have a similar problem when the first connect the probe, so I'd like to help you get this resolved, not only for your benefit, but for my knowledge as well.
Thanks Andy!