Unless you control the environment for a device under test on your workbench, just do not connect the probe ground lead. Even better, temporarily remove it from the probe so that it is not flailing around where it can contact something dangerous. For most power line work, high frequencies are not involved so the probe ground is not needed especially if the bandwidth is deliberately limited. Some oscilloscopes have bandwidth limit settings below 20 MHz like 5 MHz or 1 MHz which can be advantageous in this situation.
For the workbench where you would normally connect the probe ground to neutral, an isolation transformer needs to be used on the device under test. Besides the low impedance difference in neutral and ground potential which can drive considerable current through the oscilloscope ground, a neutral failure would drive all of the current through the oscilloscope ground. I once watched a probe ground lead turn white hot and melt when this happened; now if I do not have an isolation transformer and need one, I hack one together on the spot.
There are special oscilloscopes intended for this application which have isolated probe grounds like the Tektronix TPS series. For them, each probe ground has to be connected and it is safe to do so. An alternative to this is to use high voltage differential probes or sometimes two oscilloscope channels in add and invert mode.