Not getting new tasks

Here is a small excerpt from the log, edited to show the critical information

2017-01-22 13:30:51.7997 [PID=28994] [send] CPU: req 0.00 sec, 0.00 instances; est delay 0.00
2017-01-22 13:30:51.7997 [PID=28994] [send] CUDA: req 2167.37 sec, 0.67 instances; est delay 0.00
....

2017-01-22 13:30:51.8321 [PID=28994] [version] Checking plan class 'FGRPopencl-Beta-nvidia'
2017-01-22 13:30:51.8321 [PID=28994] [version] parsed project prefs setting 'gpu_util_fgrp': 0.330000
2017-01-22 13:30:51.8321 [PID=28994] [version] Peak flops supplied: 4.32256e+11
2017-01-22 13:30:51.8321 [PID=28994] [version] plan class ok
2017-01-22 13:30:51.8322 [PID=28994] [version] Best version of app hsgamma_FGRPB1G is 1.18 ID 936
                                                                           FGRPopencl-Beta-nvidia (108.84 GFLOPS)
2017-01-22 13:30:51.8327 [PID=28994] Only one Beta app version result per WU (#270777670, re#1)
2017-01-22 13:30:51.8328 [PID=28994] [send] [HOST#12218186] [WU#270777670 LATeah0009L_940.0_0_0.0_]
                                                                           WU is infeasible: Project-specific customization
2017-01-22 13:30:51.8332 [PID=28994] Only one Beta app version result per WU (#270780782, re#2)
2017-01-22 13:30:51.8338 [PID=28994] Only one Beta app version result per WU (#270780784, re#3)
....

The first two records show that your BOINC client is not asking for CPU work and only a small amount of GPU work - 2167.37 seconds worth.

The next block of records show the scheduler working out that it wants to send you work for the beta test version (1.18) of the gamma ray pulsar search.  It also shows you have set your GPU utilization factor for this type of work to 0.33 - meaning that you wish to crunch 3 tasks concurrently.  Finally it tells you that there are no feasible choices for a task to send to you because it is only allowed to send one beta test task per workunit and all the ones it was able to check already had that one test task allocated to other hosts.  This is a project specific customization designed to ensure that beta test tasks are validated against known 'good' tasks so as to prevent a potentially buggy test app being validated against itself.

There are obviously too many hosts requesting beta tasks at the moment :-). 

No, it's not OK.  All the searches mentioned in it no longer exist so it's really just being ignored.  The only one for which there might be work at a future date is BRP4G but even that is rather doubtful.  You should delete the file as it really is no longer relevant.

If you are going to use an advanced and highly specialized feature like app_config.xml, you really need to pay attention to the various searches as they come and go.  You also need to understand the documentation, so you can properly maintain the file as things change.  You don't need this file just to run multiple concurrent GPU tasks.  You can do this already with GPU utilization factor in your preferences - as you already have done for FRGP style GPU tasks with the 0.33 value mentioned above.  One of the reasons you might consider app_config.xml is if you needed to change the default values for CPU utilization (<cpu_usage>) that come with a particular GPU utilization factor.  Values defined in app_config.xml will override default values.

If all the above seems somewhat confusing and perhaps even rude, please be assured that neither is intended.  The server logs are a valuable resource for those who need to use them so I'm just trying to analyse examples as they come along.  Other readers may derive benefit from this.  People who want a 'set and forget' type of experience should use the defaults as much as possible and avoid the advanced stuff.  Avoiding test tasks is also in that category.  Sure, it's great to be using the fastest app, but what about when things go wrong?  The test app will quickly become the standard app if it really is more efficient and trouble free.