Step 1: Use elimination by checking each claim against a known geodetic fact.
Instead of starting from definitions, test each statement and cross it out if it fails a known result.
Step 2: Cross out option (C).
Geoid undulation $N$ (height of geoid above the ellipsoid) ranges roughly from about $-107$ m near Sri Lanka to about $+85$ m near New Guinea. Since $N$ takes both signs across the globe, the ellipsoid cannot be always below the geoid. Option (C) is eliminated.
Step 3: Cross out option (D).
A long-term tide gauge average defines local mean sea level, which is a height datum, not the geoid. The geoid is a globally consistent equipotential surface derived from gravimetric, satellite altimetry and gravity-mission data, so calling a single tide-gauge time average a realization of the geoid is inaccurate. Option (D) is eliminated.
Step 4: Cross out option (B).
National height datums have historically been fixed using a single benchmark tide gauge. A minimum of two gauges is not required for estimating a mean sea level value at a place. Option (B) is eliminated.
Step 5: Confirm the remaining option (A).
The classical Gauss-Listing definition ties the geoid to mean sea level only as a best least-squares fit; the actual sea surface differs from the geoid because of sea surface topography, so MSL approximates the geoid but is not identical to it.\[ \boxed{\text{Option (A)}} \]