ISO C17 (ISO/IEC 9899:2017 - N2176 working draft)

7.21.5.3 The fopen function
...
6 Opening a file with append mode (’a’ as the first character in the mode argument) causes all subsequent writes to the file to be forced to the then current end-of-file, regardless of intervening calls to the fseek function. ...
...

ISO C23 (ISO/IEC 9899:2024 - N3220 working draft)

7.23.5.3 The fopen function
...
6 Opening a file with append mode (’a’ as the first character in the mode argument) causes all subse-quent writes to the file to be forced to the then current end-of-file at the point of buffer flush or actual write, regardless of intervening calls to the fseek, fsetpos, or rewind functions. ...
...

[wg14/wg21 liaison] fopen 'x', 'a' and 'p'

From: Niall Douglas <s_sourceforge_at_[hidden]>
Date: Fri, 27 May 2022 13:23:43 +0000
...
fopen("a"):

Opening a file with append mode (\code{'a'} as the first character in the mode argument) causes all subsequent writes to the file to be forced to the current end-of-file at the point of buffer flush or actual write}, regardless of intervening calls to the \code{fseek}, \code{fsetpos}, or \code{rewind} functions. Incrementing the current end-of-file by the amount of data written is atomic with respect to other threads writing to the same file provided the file was also opened in append mode. If the implementation is not capable of incrementing the current end-of-file atomically, it shall fail setting \code{errno} to \code{ENOTSUP} instead of performing non-atomic end-of-file writes.} In some implementations, opening a binary file with append mode (\code{'b'} as the second or third character in the above list of \code{mode} argument values) may initially position the file position indicator for the stream beyond the last data written, because of null character padding.

[Main change: increment of end of file to become atomic]
...

Although the C17 draft (N2176) strictly mentions only fseek. The C23 draft (N3220) and the liaison material explicitly broaden the wording to include fsetpos and rewind, making the specification unambiguous across all File-Positioning Functions.

And despite the fact that the liaison material states that the “Main change” annotation highlights the atomicity addition, the wording change is undeniably present in the final proposal that WG14 moved forward with.

So finally, the mystery has been solved many years later.