We've sort of found the problem - Sybase server need to be "turned off and on again".
That fixed it.
There was a stack trace earlier in the day... not sure if its any use - but this what we got.
00:15:00000:00318:2014/12/18 10:44:01.08 kernel Current process (0x66ba01d5) infected with signal 11 (SIGSEGV)
00:15:00000:00318:2014/12/18 10:44:01.09 kernel Address 0x0x000000000199ef8b (open__byname+0x4fb), siginfo (code, address) = (1, 0x0x0000000000120028)
00:15:00000:00318:2014/12/18 10:44:01.09 kernel ************************************
00:15:00000:00318:2014/12/18 10:44:01.11 kernel SQL causing error : e
00:15:00000:00318:2014/12/18 10:44:01.11 kernel Current Procedure:
00:15:00000:00318:2014/12/18 10:44:01.11 kernel Current statement number: 1 Current line number: 1
00:15:00000:00318:2014/12/18 10:44:01.11 kernel ************************************
00:15:00000:00318:2014/12/18 10:44:01.11 server SQL Text: insert #proxy_ntuple
00:15:00000:00318:2014/12/18 10:44:01.11 kernel curdb = 25 tempdb = 7 pstat = 0x10000 p2stat = 0x40901000
00:15:00000:00318:2014/12/18 10:44:01.11 kernel p3stat = 0x800 p4stat = 0x80 p5stat = 0x8 p6stat = 0x11 p7stat = 0x10000
00:15:00000:00318:2014/12/18 10:44:01.13 kernel lasterror = 0 preverror = 0 transtate = 1
00:15:00000:00318:2014/12/18 10:44:01.13 kernel curcmd = 296 program =
00:15:00000:00318:2014/12/18 10:44:01.13 kernel extended error information: hostname: login:
