Logic:

1. Current EL load is 1M5 events a day; the system can handle the load where it is now, but reducing would be a good idea.
2. VE is ~ 70K edits -> @ 100% ~ 150K events a day
3. WT is ~ 2M5 edits -> @ 25% ~ 1M25 events a day
4. Reducing the sampling rate from 100% on VE would be irritating for bug tracing (as often these bugs only strike once or twice).
5. As the WT data is so vastly huge in comparison, cutting that even slightly would have much more impact on EL load than doing so for VE.
6. We don't use the WT data the same way, so reducing that won't be problematic from the Editing department's POV.

Consequently, drop WT sampling from 4/16 (25%) to 1/16 (6.25%) using the existing system (which uses the first hex character of the session id) as that should continue to be evenly distributed on user activities.