Can't ORFS be of any size?
I think the point being made is that in a random sequence of 64 amino acids, on average 3 out of 64 should be stop codons, because there are three possible nucleotide sequences for stop codons (UAA, UAG, UGA). ORFs could theoretically be any size but the longer the sequence the more likely you are to come across a stop codon purely due to probability.
What is overlapping frames here?
For instance, you can see that the first "CU" in the first frame produces an alanine (as GCU), while the "CU" in the second frame produces a leucine (as CUU).
They have the same nucleotide sequence, but are offset to produce different amino acids. Viruses have very "condensed" genomes and often can use the same nucleotide sequence for different genes, just starting at different points in that nucleotide sequence. This means they have functional overlapping frames.