NANO internal brownout circuit reliability

A NANO based project of mine operates on 12VDC, from a typical analog supply (9VAC transformer, rectifier bridge, capacitors). For timekeeping purposes, I want the system to have some reasonable immunity to the typical power interruptions we get around here. So I want to implement a simple backup with NI-cads. For example, I could have have 4 ni-cads feeding the 5V through a reverse blocking diode, with a resistor connected from the 12V supply directly to the NI-cads. It will keep them charged, and the NI-CAD voltage will never be high enough to feed back into the NANO 5V pin, but in the event of power loss it will take over immediately. Of course I do have "loss of power" detection in my code, so that I can quickly stop supplying power to current hogging LCDs, LEDs, etc.

Anyway, there are a few variations on the scheme I described, but my question is about how the processor will react if left running on the NI-cads for too long. With a fully charged pack of 4 NI-cads, the board will still only see 4-1/2 volts tops, and I already know the NANO will still operate at that voltage. But what happens if this goes on, unattended, until the available voltage crosses the brownout detection logic (probably somewhere around 2-1/2 to 3 volts?). I would like to think it gracefully shuts down the board, based on the spec. But batteries are funny. If there's no load after the brownout shutdown, the available voltage could jump up, enough to be in the "good enough to run" zone again. Is there a healthy amount of hysteresis in the brownout detection system to prevent such flip-flopping?

PeterPan321:
Is there a healthy amount of hysteresis in the brownout detection system to prevent such flip-flopping?

The voltage levels and amount of hysteresis are described in the 328 datasheet section titled...."Brown-out Detection."

You're looking in the wrong end of the microscope. You said early on:

For timekeeping purposes

If you're trying to keep track of time and you don't have an RTC chip with a coin cell backup battery that keeps time no matter what happens on the AC side, you're doing it wrong.

Besides, Ni-cds are so nineties eighties seventies.. Nasty things they are. Obsolete too. If you must use a battery, use a li-ion. Cheaper, smaller, higher power density and you can buy a charger for less than a buck (TP4056).