For theoretical calculations, discharge lamp voltage is classically modeled as trapezoidal.
Pretty sharp trapezoid (the top picture), I would call that mora a rectangle. Definitely the voltage shape is way closer to a "true" rectangle than the current to the sinewave. But regardless how you call it, there is clearly very significant shape mismatch, rsponsible for the lower than unity power factor. Exact rectangle vs exact sinewave would give the 0.9, square vs triangle (as the current is not far from that either) would yield about 0.866.
What I would call trapezoidal, it becomes with a resistive ballast (the middle picture). It is related to the phase shift between mains voltage vs current zero cross, but once the mains voltage at the current zero cross is higher than the arc voltage, it reignites immediately so the shape is rectangular. At least for fluorescents (where the ionization density responds pretty fast to the current).
And for the swirling: Regardless of the mechanism, I see it as very common behavior when first switched on after long (months) and cold (below 10degC) pause, disappearing after the lamps fully warm up and stabilize and uses not to appear when the lamps are used normally (at least turned on for about half hour every other day) until the lamps are left off for the months at cold temperature again. So I considered that pretty normal behavior for such abnormal use pattern.
