The tolerable misalignment is set by the beamwidth and the link margin, not by a single angle. A larger aperture gives a narrower beam and more gain, so it tolerates less sway for the same throughput; a smaller array with a wider beam is more forgiving but starts with less margin. Work out the 3 dB beamwidth for your array and the fade margin at the operating distance, then translate sway into dB: a misalignment of about one quarter of the beamwidth typically costs 1-2 dB, which a link with 6-8 dB of margin absorbs, while a link running near threshold will drop modulation order first and then lose the link entirely.
Separate slow drift from fast gust motion. Slow drift from mount creep or temperature can be handled with periodic re-alignment, but gusts need a tracking loop fast enough to follow the pole motion. Shorten the channel-state feedback interval, increase the beam-tracking rate, and keep a wider fallback beam that the system can switch to when the narrow beam loses lock. A wider beam with lower gain and a stable link is better than a narrow beam that keeps dropping out.
Mechanical fixes still matter, but they are not a substitute for tracking. A stiffer mount, a shorter pole, damping at the base and a low centre of gravity all reduce sway amplitude, and they reduce the rate at which the tracking loop must work. Measure the result in the field with a simple accelerometer or inclinometer on the pole top, logged together with beam index, received power and throughput. That log is what separates a wind problem from a rain-fade or interference problem.
In testing, compare tracking on and off under the same wind, and record whether throughput falls by modulation-order step-down or by a sudden link drop. The step-down case points to a margin and beamwidth problem; the sudden drop points to a tracking-loop or re-acquisition problem. Fix the one you actually measure.