I should point out that even when plants are in pots with "processed/manufactured" growing medium/"soil" (presumably as uniform as possible) and grown in environmental chambers where light and temperature are controlled (also presumably as uniform as possible) both randomization and replication are still important for tests/comparisons. In those circumstances the positions of the pots are usually rotated at random at set intervals.
Also the plants at the edges of a block of one cultivar have neighbours that differ from the plants in the centre of the block. The plants on the edges, called guard rows or border rows are usually not included in the measurements due to the possibility that they are different due to differences in competitive effects between neighbours of different cultivars (and due to other potential edge effects).
There are nine individual clones/divisions of cultivar A (labelled A1..A9). The other plants in the location are unknown cultivars simply labelled X. Individual clone number 5 of cultivar A (A5) has eight nearest neighbours and all are of cultivar A. Individual clone 1 of cultivar A (A1) has eight nearest neighbours but only three are of cultivar A. Individual clone 2 of cultivar A (A2) has eight neighbours but five are of clone A.
In a normal test using a block design there would be more rows and columns of cultivar A and the plants of cultivar A on the outside edges of the block would not be measured due to possible edge effects.