[nhibernate-development] Locking within QueryPlanCache

[Steve Strong]

(another) quick question:  In the various caches used for holding Query Plans, 
there's lots of code like this:

                public object this[object key]
                {
                        [MethodImpl(MethodImplOptions.Synchronized)]
                        get
                        {
                                lock (SyncRoot)
                                {
                                        object result = softReferenceCache[key];
                                        if (result != null)
                                        {
                                                strongReferenceCache.Add(key, 
result);
                                        }
                                        return result;
                                }
                        }
                }

Is there any good reason for the two locks?  Seems to me to just be unnecessary 
overhead, but I'm perhaps missing something subtle...

Cheers,

Steve