ChessBase 10 cannot handle large databases!

I'm running ChessBase10 under Win XP Home SP3.
 
1) I made a massive CBH file (7.3 million games) by merging Mega2009 database with a large, free database I found online. The merge worked fine.
 
- Then I asked CB10 to find doubles - it found at least 1.5 million doubles. It crossed these out in grey, ready for deletion.
 
- Then I right-click on the database symbol > Tools > Remove Deleted Games. Then it asks me whether I want to "Pack games"? There is no choice but to press Yes. Then CB10 performs an Integrity Check of the database.
 
It runs but soon I get a message saying "The database contains data problems. Please run Integrity Check to solve these problems".
 
But when I run the Integrity Check on the database, it says it is OK!
 
So CB10 does not let me delete the doubles!!
 
2) I tried to remove all doubles and games less than 5 moves from Mega2009 - CB10 crashed, it could not complete the Integrity Check and thus could not delete the marked games... (why are there such useless games in a commercial database anyway??)
 
3) I downloaded the new games for 2008 and 2009 to add to Mega2009 - and tried to remove all doubles and games less than 5 moves - CB10 crashed, it could not complete the Integrity Check and thus could not delete the marked games...
 
4) After copying about 4 million games from Mega2009 to a new database (all games with 5 moves or more), CB10 crashed - runtime error
 
My conclusion: CB10 cannot handle large databases, not even the ones sold commercially by ChessBase!

I've sent a (slightly angry) letter to ChessBase about this. Their products are not cheap...
 

Can you do the merge on the fly and have the doubles taken care of as the merge is occuring?  I thought that was an option... Copy mega to a new database, and then import the large free db into it with all the options enabled for skipping doubles (slow, don't care about names, keep annotations, keep best game, etc).  This is off the top of my head, so it may not be exact.

I have heard about problems with scid too when it comes to very large databases... I haven't read anything on Chess Assistant to know if they have similar issues either.

Not with recent versions (3.7 and bigger)

Chessbase is very finicky about the format of the games that are imported.  If there are any games with blanks in the player names or results or any bounded text files, it will choke mid-merge, and the file will end up being corrupted.  Corrupted files can be salvaged to some extent by exporting them to PGN, deleting the text files or games with blank headers, and then reconverting back to cbh.  In general, you should save back-up copies of any files before you attempt merges like this.  This is just an impression, but Chessbase 8 seemed fairly stable, but Chessbase 9 and 10 seem a bit buggier.

Thanks, Gonnosuke.  Good overview.  Another chessbase file I occasionally find is NAME.FLAGS which I think contains the national flags of the countries associated with the players in each game.

Another problem is some games may contain bad PGN.  I've heard that opening the PGN file in SCID, and re-saving it can fix some of these types of errors.

Also the Search Booster for large files easily gets corrupted, so you may need to right-click on the database, choose Delete Search Booster, and then Create Search Booster.