Uncomplete datasets?

grazie mille 

...there is a big gap between 9,916,896 and 10,000,000

You're right. Currently that's the largest gap (84,103 blank pages) between two tconsts on .

As of today's afternoon the highest tconst was tt10136648 whereas there were 5,769,583 unique titles, so around 43% of the tconsts are unoccupied. Well, that's only partially true because some of them can actually redirect to a higher tconst after a merging, so are not really "empty". However, since all titles added since a date some years ago have even tt numbers (from tt2404814 onwards) most of them are in fact blank title pages. Moreover, the most frequent distance between two tconsts is 2 (gap=1; 56.7%) while only 2,212,421 tconsts are consecutive to the previous one (gap=0; 38.3%).

That also explains that larger-than-1 gaps between tconsts (as a result of delete/merge processes) are mostly odd numbers (even[hi]-even[lo]-1=odd), as can be seen in the following graph (note the log-10 scale for the Y-axis):

The gap leading up to the switch to eight digit numbers may be because the sequence that allocates numbers was updated manually at a time when the relevant staff would be around to make sure there were no issues when the tconsts became longer. I suspect that the software team and the data editors had a carefully planned event where the sequence was updated and then the flow of the incoming titles was watched very carefully to ensure that everything was working correctly.