Can you tell me which part is unstable? While Im solo mining blakecoind likes to time out sometimes.
The current problem is with the shares submission to the database at some point the socket(db connection) is locking up and creating a broken pipe, this looks like it is due to the structure of a few functions that opens the connection and then opens the cursor but does not close it correctly, best practice says to close the cursor and db connection after you have finished insert but the pool code is not doing that it even reuses the socket which can also create issues and it should be flushed before reuse?
edit:
on windows due to a bug in the boost library you can get a timeout/lock that is due to be fixed in 0.9.0 by upgrading the boost library to 1.55