Submitted by: @pavel-zotov

Attachments:
gdb-pausing-fb_smp_server-and-fn_lock_print-two-snapshots-with-a-then-with-l__-n-switches.7z
shut-active_25.7z

After database state is moved to shutdown, some of ISQL sessions (which do intensive DML at this moment) can return control to OS extremely slow, elapsed time is measured about 5-7 minutes.
Batch + .sql scripts are in attached .zip.

When this occured, I've ran:
1) gdb (without arguments) for suspending FB process,
2) made two "snapshots" of lock table (with "-a" and "-l -n" swicthes), and finally
3) obtain remaining stacktrace by issuing commands in gdb prompt:

set logging on
set height 0
thread apply all bt

Lock-print data and stack trace see also in attached file.

====== Test Details ======

2do: return to core_5087.fbt and assign there min_version = 2.5.6 after this ticket will be fixed.