Investigate "still reachable" valgrind memory errors

Most of the valgrind logs are full of:

==9548== 4 bytes in 1 blocks are still reachable in loss record 1 of 51
==9548== at 0x4027D17: operator new(unsigned int) (vg_replace_malloc.c:292)
==9548== by 0x4E633A8: google::protobuf::DescriptorPool::DescriptorPool(google::protobuf::DescriptorDatabase*, google::protobuf::DescriptorPool::ErrorCollector*) (descriptor.cc:770)
==9548== by 0x4E6346D: google::protobuf::(anonymous namespace)::InitGeneratedPool() (descriptor.cc:816)
==9548== by 0x546E8DF: pthread_once (in /lib/libpthread-2.12.so)
==9548== by 0x4E86E3F: google::protobuf::protobuf_AddDesc_google_2fprotobuf_2fdescriptor_2eproto() (descriptor.pb.cc:656)
==9548== by 0x4E8790C: __static_initialization_and_destruction_0(int, int) (descriptor.pb.cc:705)
==9548== by 0x4ED2495: ??? (in /home/hudson/workspace/blacktie-linux32/nbf/target/cxx/runtime/lib/libprotobuf.so.7)
==9548== by 0x4E4538C: ??? (in /home/hudson/workspace/blacktie-linux32/nbf/target/cxx/runtime/lib/libprotobuf.so.7)
==9548== by 0x400EFCE: _dl_init (in /lib/ld-2.12.so)
==9548== by 0x400088E: ??? (in /lib/ld-2.12.so)

Please can you explain what this is and whether it is a leak and if so, try to resolve it?