Here is a rough and ready approach, I am sure you could find something more rigorous in the scientific literature.
Elevation of noon sun varies +/- 23.5 degrees, for S Coast this means 16 deg (winter solstice) to 63 (summer solstice) - see
Sun path - Wikipedia, the free encyclopedia. Like day length, this variation with the seasons will be approximately sinusoidal with half of the total change happening during the periods 30 days either side of the equinoxes (Mar 20 and Sept 22).
So this would suggest altering the tilt of the array on 18 Feb, Apr 17, Aug 25, Oct 20 in 11 3/4 deg steps, which will ensure the pitch error is never more than 11 3/4 deg from the optimum. However it does not allow for the fact that near the solstices the noon elevation changes only slowly so more time is spent with errors. Correcting for this would give you more like 15 deg steps 45 days either side of the equinoxes i.e. 3 Feb, 2 May, 10 Aug, 4 Nov.
Doing this will reduce the maximum error from ideal pitch to +/- 15 deg and fits neatly with your plan to change pitch four times a year with a range of 30 deg. [I would have thought the actual values you suggest (30 to 60 deg) are a bit high though, even for the S coast, but see below.]
Assuming the array is facing due S, looking at the Orientation and Pitch Factor table (e.g. the draft DTI Guide Ed. 3 section 3.7.3.1) the pitch factor for a year-round pitch error of 15 deg is 0.97. (Cosine of 15 deg is .0966). However since most of the summer the error will be <10 deg we should perhaps use the figure for 10 deg which is 0.98
In contrast the pitch factor corresponding to 25 deg year-round pitch error is 0.93 (with a fixed array this is approximately the error in mid-summer or mid-winter). So it looks as though if you adjusted the elevation four times a year there could be at most a 5% improvement. Allowing for the fact that a fixed array is not too far from the optimum for maybe half the time anyway, an improvement of half the difference might be attainable in practice i.e. 2.5%.
Ideally we would also make a small adjustment for the fact that summer daylight is more intense and the hours are longer, this would suggest moving all the dates even nearer to the summer solstice and increasing the elevation slightly but I don't immediately have a method for estimating this.
HTH
