I'm assuming Dapper doesn't do any pre-sizing of the list, yes? I ask because it's a requirement of the test that it doesn't.

If you mean "to match the expected data": then no, dapper doesn't do that, because it is consuming a raw IDataReader (DbDataReader on core-clr) API, and doesn't know the number of rows until it reaches the end. In this setup, it is essentially:

List<T> buffer = new List<T>();
while (await reader.ReadAsync(cancel).ConfigureAwait(false)) {
    buffer.Add(func(reader));
}
while (await reader.NextResultAsync(cancel).ConfigureAwait(false)) { }
return buffer;

(where func is a delegate that processes each row into the target data type)

Reference source is here; note that command.Buffered is true by default (most people screw up with things like concurrent commands, if handed back an open reader - so this defaults to the safest option of consuming the data before handing it back to the caller)

I ask because it's a requirement of the test that it doesn't.

Do you have a link to the test requirements? I would be happy to check them all.

https://www.techempower.com/benchmarks/#section=code

Thanks @benaadams ; I have reviewed the requirements; dapper does nothing that would violate Test 4 rule 5, if that is the concern.