2048 in fetch_all_by_dbID_list():

  # Ensure that we do not exceed MySQL's max_allowed_packet (defaults to
  # 1 MB) splitting large queries into smaller queries of at most 256 KB
  # (32768 8-bit characters).  Assuming a (generous) average dbID string
  # length of 16, this means 2048 dbIDs in each query.

The observed decrease in time for unsorted dbID lists of 10000 IDs
ranged between 5% to just under 10% depending on feature type, lower
(around 3%) for sorted ID lists.

Some reformatting.