While I'm preparing manual entries for srf-121, I felt I need to add the following

description to make-for-each-generator:

If obj is mutated before the returned generator walks all the values, the

behavior depends on how the for-each procedure handles the situation;

it may or may not be a safe operation. In general it's better to avoid

mutation until the generator returns EOF. Once the generator is exhausted,

though, it is safe to mutate obj.

Is this worth including to srfi document? It may be implied, but

how mutation interacts with already exhausted behavior may not be obvious.