Excessive CPU usage of 4s-httpd on SPARQL Update about 4store HOT 1 OPEN

I'm cross-posting this from the message group in case it's more useful here.

I don't believe the problem is isolated to DELETE or SPARQL. I am evaluating 4store as a graph database, and I made a python script that generates 15,600 triples and POSTs them to 4store in Turtle format. I first create a graph with 6 triples, then I append to it (POST) in various ways as a test:

1. Join the turtle triples with newlines, and send one request with a graph.
2. Send each individual triple with a new HTTP connection and request.
3. Reuse HTTP connections, but still do a separate POST for each triple.

Results:

Took 0.71 seconds to batch import 15600 triples
Took 400.48 seconds to concurrent import 15600 triples
Took 271.22 seconds to concurrent import 15600 triples, reusing connections

In cases 2 & 3, 4s-httpd is using 95% CPU the entire time. Since I want to use 4store as a graph database for an application, my usage model will be closest to the 3rd scenario (lots of one-shot requests with a single triple). The performance overhead with the separate requests is astounding; hopefully there's a low-hanging-fruit bug to fix it.

EDIT:
I should mention I'm on a 2010 Macbook Air running the default configuration of 4store, and I used 10 concurrent threads (actually greenlets) to do the POSTing in cases 2 and 3.

from 4store.