MTime will generate a time with micosecond accuracy if the file system supports it. This can cause problems of coherence between machines where the copy mechanism is not more accurate than 1 second. This could be fixed if FileETag had an option MTimeSeconds that produced MTime with a 1 second granularity. The fix is not hard and would involve a few lines in: ap_make_etag() in http_etag.c set_etag_bits() in core.c
This is quite important issue and the proposed MTimeSeconds must be added because now there is no ability to configure Apache HTTPD to return the same ETag as Nginx. Also I did a research how other web servers generates ETag and my proposition is to make Apache by default return the same ETag as Nginx i.e. MTimeSeconds-Size. I asked a question in http-wg to add to spec a recommended way of ETag generation "by default" https://lists.w3.org/Archives/Public/ietf-http-wg/2020JulSep/0041.html Please leave your thoughts and opinions in that discussion.