> 'pool' and 'journal' answered my questions. thanks for the fast response!
Yes, 'pool' is the right place.
Please note that 'journal' has strictly speaking not much to do with
ACID. It writes a protocol of changes (a journal file) to some other
database, and is used for life-mirroring a database (typically with
"bin/replica") on a remote machine.
The relevant argument to 'pool' here is 'sym3'. If given, a file with
that name is maintained during 'commit', into which all data modified
during a transaction are written first, then secured with fsync(), then
the data are actually written to the db file. If the process is killed
(e.g. with -9) during that process, no data will be lost, as at the next
start (call to 'pool') and automatic rollback will be done. Thus, no
other function like 'journal' is involved here.