Hello and sorry for tis late response,
yes, i have tried out "Tell HTTP quit". The result is, that every minute the alert "waiting for all tasks to quit" came on the console. The http service terminates at first, but the task runs and sends these alerts out. And this state does not stop. "Load HTTP" does not work in this moment because of the not really terminated HTTP-Task. The only way to terminate this situation is "restart server" - and also in this case i have to wait for the timeout in 5 minutes.
Now i have tested and tryed out a lot and in this moment the HTTP runs since 4 days without any error. I am very sure, that it was a defect (and big) fulltext index. Search queries to this index are sent in our system from a web triggered agent. Normally this index has a size about 1.2GB. As i looked at it, the size was 1.5GB. In recent years i have often found errors in this index. As we know, indizes with big size are very fragile. This is the reason, why i scheduled a updall -X on it every night. But this scheduling - it seems so - does not repair this. The only way is to delete it manually and set it new... :-(
Hope that it works now....
Thanks, Bernd