If you mean 4096-byte (4 KiB) chunks, yes, that seems pretty normal. (Writes to stdout and elsewhere can be buffered by the underlying library, and the default can be either “buffer up to the next newline” or “buffer in larger chunks” depending on whether stdout is a tty or not.)
I can think of a couple of differences between images (many are based on the Alpine distribution, which has a different libc; the specific ping(1) binary may or may not call setvbuf(3) or fflush(3)). Of course, Docker suggests all of the MongoDB images are Debian-based, so…?