How about a max_insert_batch method ?

Hi,

I deal with lots of data, that is several 100k rows x several websites every day.
Data is not always clean (external origin), so insert_batch sometimes does not insert all of it, and it's ok.
But i'd like to maximize the amount of data really inserted in database, so i was thinking about extending DB_query_builder to add a "max_insert_batch" method, which would do almost the same as insert_batch, except this :
each time you call _insert_batch, you verify the number of affected_rows
  if it's different from the number of lines sent (which is batch_size, except for the last batch), you recursively call insert_batch with the same block of data you just used, but with a batch_size twice as small (unless of course batch_size = 1)

Your opinion ?

Thanks