An Ohmmeter measures resistance by applying a current to its probes, and measuring the resulting voltage. As a result, in any circuit which generates its own voltage, you will have the potential for errors.
In the case of 2 back to back batteries, that's a good try. However, any slight difference between the voltage of the batteries (and there will be, battery voltage changes with discharge state, discharge history and temperature) will give a non-zero voltage.
When measured one way, this voltage will add to the resistance*current voltage, and so artificially increase the apparent resistance reading. When measured the other way, it will subtract, and may even give a negative voltage. A meter will not have been designed to interpret a negative reading, and so will probably read zero.
The only reliable way to measure the internal resistance of a battery is to measure changes in terminal voltage when you change the terminal current, either by changing the load, or using an AC excitation signal.
When using a meter to measure resistance, a different reading for either polarity is a good indication that 'something is going on', and neither reading should be trusted. Badly behaving circuits could include (not an exhaustive list) a battery, a diode or other semiconductor, a big capacitor with some residual charge on it, or junctions with thermo-electric voltages being developed.
With experience, you can use the fact that a DMM on ohms is a current source with a voltage measurement, and use it to investigate components other than resistors. The rate at which the reading changes when measuring the 'resistance' of a large capacitor can give you some indication of its value. Be aware that different ranges will use different currents, and an auto-ranging meter will quite happily switch ranges when you don't expect it.