Comment by sfkeller

Here's an implementation with lines of sql and pl/pgsql:

Given a table mytable with an pk id and say the fields 'block' and 'hash', replace it with mytable_changes and add an "internal" field called created_at of type "timestamp not null default now()".

Then create this trigger function (e.g. PostgreSQL):

  create function mytable_changes_trigger() returns trigger as $$
  begin
    if (tg_op = 'insert') then
      insert into mytable_changes (block, hash)
      values (new.block, hash);
      return new;
    if (tg_op = 'delete') then
      insert into mytable_changes (block, hash)
      values (old.block, old.hash);
      return old;
    elsif (tg_op = 'update') then
      insert into mytable_changes (block, hash)
      values (old.block, old.hash);
      return new;
    end if;
  end;
  $$ language plpgsql;

And create this trigger:

  create trigger mytable_changes
  after insert, update or delete on mytable
  for each row
  execute function mytable_changes_trigger();

What's left as an exercise is that the hash of the prev. block is being preprended.