Hello there,
To test a cap for leakage you apply a voltage and measure the current. But the current could be low so you need a low current meter to be able to measure it. You can apply different voltages by starting at a low voltage like 5v, then measure the current. Then turn the voltage up a little at a time and see that the current does not increase too much. You'll have to decide for yourself what is too much leakage current in the given application. It could also be that the leakage increases with temperature so keep that in mind.
The leakage current acts to change the charge and discharge time as Eric pointed out. The difference is that it acts like an equivalent resistor that is put in parallel with the charging or discharging resistor. So the timing itself changes because the equivalent resistance is acting in parallel with the normal charge resistor. The approximate equivalent resistor is equal to the voltage divided by the leakage current, where this voltage would be the average voltage over the charge period as an approximation.
Timing circuits with RC elements are always subject to variations in timing due to temperature anyway though. So for best results you have to keep the circuit at a relatively constant temperature for good results.
As others have pointed out, a single stage RC is not a good idea for very long time delays, although coupled with the 555 timer IC it works to some extent. But better is the faster oscillator and digital divider circuit made using counter IC chips like the 74LS93 for example. With this one chip you can make the oscillator 16 times faster and still get the longer time delay. With two chips you can make the oscillator 256 times faster, etc. So you can get some VERY long delays like this.
You can also get chips that are made just for this purpose as they contain many divider stages inside. So you can see very large division ratios like 1000 or something.
This is also a technique used for some of the older TTL frequency counters. They divide down the output frequency of a crystal oscillator and end up with a 1 second gate. So that's a division of over 32000. That would clearly be a good way to develop a time delay that needs to be accurate.
Of course if you can use a microcontroller you can do this in code. You then have the option of using an external crystal or using the internal RC oscillator, but this RC oscillator is compensated to some degree so it not your typical RC oscillator made by the hobbyist. It is also calibrated at room temperature, and you can even use software to calibrate it yourself. The typical variation is about 1 percent over a fairly wide temperature range which is much better than a stand alone RC oscillator.