Assuming that you are already handling exceptions at the edges of the system, i.e. knowing that (-179.9999 ; 0) is as close to (179.9999 ; 0) as (-179.9997 ; 0).
Assuming also that you handle the rounding of decimals. For example, the point (0.00019 ; 0) is closer to (0.0002 ; 0) than to (0.0001 ; 0).
Assuming also that you understand that the distance between two points that differ by the same number of (decimal) degrees in one or both coordinates is not the same. That is, the point (0; 0.00016) is closer to (0; 0) than the (0.00014; 0.00014).
The only drawback I see is that the definition of "how close one point will be to the other" is variable. For example, the distance between (0 ; 89) and (1 ; 89) is extremely different from the distance between (0 ; 0) and (1 ; 0). Although both have the same difference in geographic coordinates (one degree of longitude).
It will not be possible to establish a distance constant value. Because the distance that separates two points of equal difference in geographical coordinates depends on their location in the sphere.
Therefore, something like "tell whether 2 points are close to each other, say at proximity of 50 meters" will not be possible without previously calculating the distance between both in terms of great-circle distance or minimum arc length of sphere between they.