Comment 9 for bug 607961

Gary and I investigated whether or not soft timeouts can be caused by
very slow clients. They can not. The soft timeout mechanism only
counts time spend in the app (not time spent sending the response to the
client or time reading the request from the client). Therefore there
must be some other reason that the request took so long to process.

We perused the various graphs of machine metrics and couldn't discern an
external reason why the response would take so long (CPU contention,
swap, etc.).

Garry suggested that we instrument the suspect parts of the code
responsible for returning the WADL with timers that will generate an
informational OOPS if any of the timed parts of the operation exceed a
given time. Two good candidates are: when generating WADL and when
reading the WADL from disk.

Also, since we don't expect the WADL to ever be generated in prodcution
(other than at build time), perhaps we should generate an informational
OOPS whenever generating WADL in production.