Laser interference pattern flickering

[basiler]

TL;DR

simple michelson interferometer setup. With path length constant the laser intensity flickers randomly. When path length changes periodically, shadows ripple across interference pattern (instead of the pattern itself changing)

I have set up a simple interferometer as follows: a laser is aligned by two mirrors and then passed through a 50/50 beamsplitter. Both arms have mirrors which reflect the beam back through the beam splitter and to a camera, which measures the pattern (concentric circles, as expected).
When both arms are stationary, the intensity of the beam all throughout the interferometer randomly fluctuates between bright and completely dim. This persists even when just one of the arms is blocked, but the beam appears stable when I block just before the beamsplitter. I thought this would be due to back reflections re-entering the cavity and destabilising it so I have added a poor man's isolator between the two aligning mirrors, but this has not improved it (I tried every angle of the linear polariser relative to the QWP, none seemed to work better).
Additionally, when one of the arm lengths is varied sinusoidally using a piezo, the pattern itself does not ripple as expected but has dark ripples over it (with a seemingly different centre and radius).

I had hoped the isolator would fix this, does anyone have any other suggestions?

Images:
TOP: the shadows rippling over the pattern when one arm length is modulated sinusoidally. I later found that the pattern was due to the laser diffracting through an aperture I had used for alignment which has since been removed, but the effect observed is still the same (the file I have of the updated interference is too big to upload)
BOTTOM: the flickering observed at the camera when both arm lengths are kept constant

 

[Cthugha]

The upper image looks like what you would see if the two beams overlap at the position of the camera, but are not perfectly parallel. How do you check the alignment of the beams?

For adjusting an interferometer, one may have a look at the two beams in real space and try to overlap them and then to have a look at the beams in momentum space (e.g. by putting in a lens - beams coming to the lens at different angles get focused to different points in the back focal plane of the lens - and overlapping the beams in momentum space as well). Repeated alternate overlapping in real and momentum space should get you good overlap in the end after a reasonable number of alignment cycles.