When the web server shuts down, it's just too much workload for how many features do you have? What does perfmon show for avg seconds read / write for a physical disk? It can be as simple as throwing money at your disk subsystem.
You might want to try this instead of reorg:
- create a new index with a different name identical to the existing index
- delete an existing index after creating a new index
- If you use query hints that specify the index name, you will need to rename the new index according to the old index name
source share