The same error happened when the rescued task was reported and new work was requested. I've done the same edits to recover the latest task and have disabled MCM work to check if my problem is specific to that application.

The system had been running OET1 with occasional tasks from other applications when that was out of work. A couple of weeks ago I switched the priority to MCM1 and all was fine until the last 2 tasks.

My other system (Q6600, XP, same BOINC version over the same router) has had no problems receiving MCM1 work over the same time period.

Guess what. The encoding problem happened again. This time I manually deleted the (already reported) completed task from client_state.xml and the new task was accepted:

03-Jul-2015 08:05:29 [World Community Grid] Sending scheduler request: To fetch work.
03-Jul-2015 08:05:29 [World Community Grid] Reporting 1 completed tasks, requesting new tasks for CPU
03-Jul-2015 08:05:32 [---] Project communication failed: attempting access to reference site
03-Jul-2015 08:05:33 [---] Internet access OK - project servers may be temporarily down.
03-Jul-2015 08:05:33 [World Community Grid] Scheduler request failed: Unrecognized HTTP Content-Encoding

03-Jul-2015 08:10:22 [World Community Grid] Sending scheduler request: Requested by user.
03-Jul-2015 08:10:22 [World Community Grid] Requesting new tasks for CPU
03-Jul-2015 08:10:26 [World Community Grid] Scheduler request completed: got 1 new tasks
03-Jul-2015 08:10:26 [World Community Grid] Resent lost task MCM1_0015047_3906_1

I've now opted out of MCM1 on that system.