I don't know if this has been brought up in the whole binary operator discussion, but the current set of binary operators are intentionally defined as experimental infix functions.
This is mainly due to the fact that the design could not be finalized yet, as it is interconnected with a whole range of things. See also
https://discuss.kotlinlang.org/t/when-does-bit-fiddling-come-to-kotlin/2249
I guess that one of the reasons to go for keywords is partly due to the technical implementation of experimental language symbols. (Maybe JB can confirm/correct?)
As far as I understood experimental functions are inlined at call site, leaving no trace of the originally called method, which allows JB to change the language feature without breaking backward compatibility.
Regarding the discussion of binary operator symbols vs keywords,
I think that readability is the most important factor to tip the decission in favor of keywords to me.
The argument that symbols might be compacter and therefore more readable vanishes in the face of the seemingly endles binary expression one liners I came across so far.
Generally I liked the point raised before by
@jw, that longer keywords encourage splitting the long expressions into several lines or functions and therefore improve readability.
In general though we should wait and see what JB will come up during their journey towards propper binary operators/keywords, as it seems like it could be a long one😀