Hi John,

> The main program should control what is contained in the sequential HTTP requests, and delegate which server should be asking what.

I'd use a job table for that. The main program inserts a new record. The servers wait for new record and attempt to lock it. Once a record is locked, the application sends the query, stores the result into another memo field, set a lDone flag in the record and unlocks it. The main application could then scan for records with the lDone flag set to process the results. If a server hangs, the main program can kill it. This automatically releases the lock and the next available server would find the record automatically.