Tracker scrape
A scrape, aka "Tracker Scrape", is a request sent by the bittorrent client to a tracker. A request is sent, connection to the tracker is established, information is exchanged, then the connection is closed. A scrape is what every BitTorrent client (such as Azureus) does, to any tracker that is hosting a .torrent which is loaded into the client. The request does something like a "wipe" or a "pass" over the tracker, and then the tracker sends information back to the client.
Please note that some trackers don't respond to scrape requests, but you will still be able to download the torrent. The returned information can contain such information as, whether the tracker is OK or offline, the reason it is offline (unknown host exception, hash missing, etc), the numbers of peers and seeds, etc.
Every BitTorrent client scrapes the tracker many times during the course of a download to update the swarm information. So you can imagine that the tracker is scraped many thousands of times for that torrent alone, even if the swarm is not very big. The tracker can usually handle this number of requests. However, if there are more requests than strictly necessary, this can destabilise the tracker and put it offline.
Why scraping?
While a torrent is incomplete, the client scrapes in order to determine whether or not to send an announce requesting more peers. Sending a list of peers is usually more bandwidth consuming than sending a scrape result.
When a torrent is complete, the client periodically scrapes in order to determine which torrents are the neediest. Without scraping, the client would never know which torrents have no seeds and require assistance.