shared memory weirdness

We're running a very old version of HPUX and Informix, the system will shortly be retired so we're limping along for a short time. We're running out of shared memory though and it's killing the database, to an extent it's something we're going to have live with but something puzzles me. From onstat -g seg : INFORMIX-OnLine Version 7.23.UC6 -- On-Line -- Up 00:56:47 -- 1077264 Kbytes Segment Summary: (resident segments are not locked) id key addr size ovhd class blkused blkfree 1030 1383417858 a0000000 440410112 7328 V 10578 43183 1031 1383417859 c0836000 606208 616 M 67 7 7172 1383417857 c0bba000 662708224 10992 R 80891 6 Total: - - 1103724544 - - 91536 43196 Which looks perfectly normal. But, from ipcs -ma : m 7172 0x52754801 --rw-rw---- root informix root informix 33 662708224 17645 18885 11:30:00 11:30:00 10:35:29 m 1797 0x00000000 D-rw-rw---- root informix root informix 33 536870912 17645 17645 10:35:29 no-entry 10:35:29 m 1030 0x52754802 --rw-rw---- root informix root informix 33 440410112 17645 18885 11:30:00 11:30:00 10:35:38 m 1031 0x52754803 --rw-rw-rw- root informix root informix 34 606208 17671 18983 11:30:47 11:30:00 10:35:49 So we have a segment, number 1797, that's pretty big, created and last accessed by the same process as the other segments, it's been removed but won't get returned to the free list until the oninit exits. So, my question is this, what's happening here? And, can I stop it doing it? Ian