As k7elp60 said, it's probably leakage, which prevents the cap from being able to charge to the pin 6 threshold. If so, the multimeter will only make the problem worse unless you measure from VCC (instead of GND) to the cap. You can pretty much get guaranteed oscillation, albeit at higher than the calculated frequency, by connecting the + end of the cap to VCC and the - end to pins 2 and 6. The leakage now aids the 4Meg resistor instead of fighting it. I doubt that you will be happy with the results, though. See schematic below.
The output will be low when you turn on power, instead of being initally high as it is with the standard cap connection.