commit 1af8bf39 added unnecessary
calls to fil_write_flushed_lsn() during redo log resizing at
InnoDB server startup.

Because fil_write_flushed_lsn() is neither redo-logged nor doublewrite
buffered, the call is risky and should be avoided, because if the
server killed during the write call, the whole InnoDB instance can
become inaccessible (corrupted page 0 in the system tablespace).

In the best case, this call might prevent a diagnostic message from
being emitted to the error log on the next startup.