Comment 1:

It doesn't seem to be related to time.  It seems to be related to two
GobEncoder/GobDecoders in the same value where one is a pointer and the other isn't:
package main
import (
    "bytes"
    "encoding/gob"
)
type T struct{}
func (T) GobEncode() ([]byte, error) {
    return []byte{0xDE, 0xAE, 0xAD, 0xBE, 0xEF}, nil
}
func (T) GobDecode([]byte) error {
    return nil
}
type PC struct {
    A     T
    B *T
}
func main() {
    buf := new(bytes.Buffer)
    enc := gob.NewEncoder(buf)
    var err error
    if err = enc.Encode(PC{}); err != nil {
        println("A", err.Error())
        return
    }
    dec := gob.NewDecoder(bytes.NewBuffer(buf.Bytes()))
    if err = dec.Decode(new(PC)); err != nil {
        println("B", err.Error())
        return
    }
}
http://play.golang.org/p/7MfiN4ToAZ

Comment 2:

It looks like the arrangement of pointer/non-pointer fields makes a difference. If the
pointer field is first, there is no problem, in other positions the bug appears.

Comment 3:

Dan is correct.  It appears to happen because GobEncoders are treated specially and as
such, the *T is encoded immediately, despite the T having already been encoded.  The
"extra data in buffer" error is actually masking the "gob: duplicate type received"
error (which we might also want to address).
I don't know if this is the proper fix, but it makes my test pass:
https://golang.org/cl/7085051

Comment 4:

This issue was closed by revision bc1152a.

Status changed to Fixed.