Schizophrenic coding in gin_extract_jsonb(_hash)

Would someone care to defend this code?

int total = 2 * JB_ROOT_COUNT(jb);

...

if (total == 0)
{
*nentries = 0;
PG_RETURN_POINTER(NULL);
}

...

while ((r = JsonbIteratorNext(&it, &v, false)) != WJB_DONE)
{
if (i >= total)
{
total *= 2;
entries = (Datum *) repalloc(entries, sizeof(Datum) * total);
}

The early-exit code path supposes that JB_ROOT_COUNT is absolutely
reliable as an indicator that there's nothing in the jsonb value.
On the other hand, the realloc logic inside the iteration loop implies
that JB_ROOT_COUNT is just an untrustworthy estimate. Which theory is
correct? And why is there not a comment to be seen anywhere? If the code
is correct then this logic is certainly worthy of a comment or three.

regards, tom lane