I was wrong about significant slowdown with
SHAREDSQL=N I mesured again and found that it is
slower but not drastically.
Also I found that setting SHAREDSQL to N didn't fix
the problem of incorrect result sets. It still happens
on 7.6.0.12. I tried again this test on 7.5.0.30 and
it passed without any problems several times in a row.
Any ideas?
Regards.
Alexei