WAL log file corrupt after

[979357361]

we use rocksdb through C API, the WAL config is kept default, underlying file system is a default EXT4 on NVME SSD

after a terminal disconnect, we try to reopen the DB and found some data is wrong, then I check and find WAL may be corrupted as LOG shows:

as far as I know, RocksDB promises data consistance with WAL enabled, and EXT4 has is own journal, too, why data still happens? Will changing WAL recovery mode to absolute_consistancy help? In our use case, we prefer no data than wrong data.