Puzzling "reduction too large" error - mapreduce

Question

I have a document bucket with the following format:
{
"a": 1,
"b": 2,
"c": 3,
"d": 4,
"e": 5
}
This is my mapping function:
function (doc, meta) {
var summary = {
b: doc.b,
c: doc.c,
history: [{
d: doc.d,
e: doc.e
}]
};
emit(doc.a, summary);
}
And here's my reduce function:
function(key, values, rereduce) {
for (i = 1; i < values.length; ++i)
Array.prototype.push.apply(values[0].history, values[i].history);
return values[0];
}
And finally, here are my view query parameters:
?stale=false&group=true&reduce=true
This combination leads to a reduction too large error. For now, I have at most 2 of the same keys in my dataset, so I can change the reduce function to:
function(key, values, rereduce) {
if (values[1])
Array.prototype.push.apply(values[0].history, values[1].history);
return values[0];
}
Keeping everything else the same, I get the exact result that I want:
{
"key":1,
"value":{
"b":2,
"c":3,
"history":[
{
"d":4,
"e":5
},
{
"d":6,
"e":7
}
]
}
}, many more like this
However, I expect the number of keys to grow over time, so I need to be able to loop over values.
I also changed my reduce function to simply return values.length in order to check if it was returning unreasonable values, but everything I get is either 1 or 2.
What is going on?
UPDATE:
When I change my reduce function to:
function(key, values, rereduce) {
if (rereduce) return "whatisgoingon";
for (i = 1; i < values.length; ++i)
Array.prototype.push.apply(values[0].history, values[i].history);
return values[0];
}
I get the results that I desire, however whatisgoingon is nowhere to be found in the view result. I don't want to "discard" the rereduce logic since I might actually use it in the future, so I need to understand the cause of the issue.