[vera-rubin]

When we write one CDF_EPOCH16 timestamp to a nonsparse variable, it becomes two records with zero picoseconds. For complex(63871286400, 123456789012), we get [63871286400+0j, 123456789012+0j] and Last_Rec=1, instead of one complex value and Last_Rec=0, as the documented representation requires.

from cdflib import CDF, cdfwrite

with cdfwrite.CDF("epoch16.cdf") as writer:
    writer.write_var(
        {"Variable": "Epoch", "Data_Type": writer.CDF_EPOCH16,
         "Num_Elements": 1, "Rec_Vary": True, "Dim_Sizes": []},
        var_data=[complex(63871286400, 123456789012)],
    )
with CDF("epoch16.cdf") as reader:
    print(reader.varget("Epoch"), reader.varinq("Epoch").Last_Rec)

We now pass EPOCH16 values directly to the shared converter, as sparse writes already do, preserving both components and the record count.

We added a regression test and ran the existing reader/writer tests.

— ChipLab