The IPI identifier tracking database "IPI History" is also available in dbfetch/WSDbfetch. So Pierre can just tweak the URLs being used to get information to find out why. Just replace the database name (db) with 'ipihistory' and either set the format to 'default' or remove the format specifier. For example:
http://www.ebi.ac.uk/Tools/dbfetch/dbfetch?db=IPIhistory&id=IPI00000178 http://www.ebi.ac.uk/Tools/dbfetch/dbfetch?db=IPIhistory&id=IPI00000495 http://www.ebi.ac.uk/Tools/dbfetch/dbfetch?db=IPIhistory&id=IPI00896455
The returned tab-delimited table details which versions of IPI the identifier appeared in, and details of why an entry was removed (see http://www.ebi.ac.uk/IPI/HistoryFormat.html). If the specific entry/sequence data referred to by the IPI is required, then you can either look-up the IPI identifier in UniParc (it helps if you have the full identifier including the sequence version), also available in dbfetch/WSDbfetch, or use the IPI release information from "IPI History" to identify the required release files on the FTP site (ftp://ftp.ebi.ac.uk/pub/databases/IPI/).
Production of the IPI database ceased last year (http://www.ebi.ac.uk/IPI/IPIhelp.html). IPI has been replaced by datasets provided by UniProt, which is extending equivalent complete proteome coverage into many other species. Where possible IPI use of IPI should be replaced by the use of the equivalent UniProt entries. For cases where you are only interested in the sequences, I suggest searching UniParc since this has all the IPI sequences and provides details of identical sequences in other data sources.