On Nov 7, 2005, at 2:37 PM, Vivek Khera wrote:

> While I wait for lock set() to return, I see that all the slony  
> processes in both origin and replica are in <IDLE> state (ie, no  
> transaction...)

for the archives:

the answer was that lock set() waits for the transaction ID's to  
advance far enough to ensure no data is missed on the new origin, so  
it was doing more than just waiting for the lock.  it was waiting for  
all old transactions to complete, which was taking a long time due to  
another DB on the old origin.