How do I recover from "Invalid log entry in replay" when transaction log is being replayed during server start?

A Virtuoso transaction log file may fail to replay, with an error of the form:

11:42:57 Invalid log entry in replay. Delete transaction log virtuoso.trx or truncate at point of error. 
Valid data may exist after this record. A log record begins with bytes 193 188 5 188 0. Error at offset 13205546
11:42:57 Searching for the next valid header signature starting at 13205547
11:42:57 No valid looking header found.
11:42:57 Server exiting

This can be recovered by truncating to the transaction log to the reported error offset.

Users of Unix-like OS can use the built-in dd command; Windows users must first install WinDD or cygwin to get the same functionality.

Given the error above, the recovery steps on a Unix-like OS would look like --

# mv virtuoso.trx old.trx
# dd if=old.trx of=new.trx bs=1 count=13205546
13205546+0 records in
13205546+0 records out
# mv new.trx virtuoso.trx
# virtuoso-iodbc-t -f

Restarting Virtuoso should now succeed.