file locking will fail with threaded access, e.g. with the apache worker mpm.

this is because a second call to apr_file_lock() from a same process (in a different thread) will fail, thus meaning that the second client request won't wait for the first one to finish rendering a metatile before going to query the cache.

for now, users should use the prefork mpm, or bear with failed tile for unseeded requests.

This issue was closed by revision r238.