This is a very common problem in my experience and the g's are very similar to mine and the waveform looks continous with no regular spikes, ie no ball pass impacts.
I suggest the cause is metal to metal rolling contact due to lack of lubrication at the rolling surfaces, not skidding, as suggested, unless you have evidence that it is.
The effect of poor lube is metal to metal contact and the surfaces micro weld together and break as they roll causing a "frosted" appearance (like pulling an arc welding rod of the job after it has stuck) which has surface roughness greater than the normal grease film thickness of no more than several polymer layers.
Greasing temporarily increases the lube film thinkness and the contact ceases. You can estimate the surface roughness by how long it takes for the vibration to return, several days is quite long, several seconds is quite short.
This type of fault is often caused when a motor is first started and the grease is squeezed out of the rolling elments very quickly, but it's too cold to "slump" back into the rolling elements and stays stuck to the housing walls, so the bearing is running dry, and the frosting begins right then. Once started it may reduce if you put a Permalube on, but will never smooth over to as new condition.
For long life it's absolutely critical to use a grease that slumps properly at operating temperature of the inside of the bearing housing. Put a dob of grease on the outside and see if it slumps ? If not use a lighter one.
The micro welds form the initital surface stress concentrations which begin to spall in time, so these bearing have a short life in terms of a few years, rather than many B10 lives.
A B10 life is the time 10% of a large group of bearings will fail within, and for 50 KW motors it's usually 3 years.
Hope this helps.
John