fdb_dump failing with"address is not freed" - possible db corruption

Updated e2e testing to avoid opening snapshots with pre-compacted handles:
http://review.couchbase.org/#/c/47512/1

test still fails. sometimes docs are missing from iterator or a fdb_get call will fail.
I tried to unpack db to see if I could find key there but ran into these errors:

../../forestdb_dump e2edb_main.1

Unable to open e2edb_main.1
address 0x00007fb8b4003200 (allocated at /Users/tmcafee/sandbox/forestdb/src/hash.cc:32, size 16384) is not freed
address 0x00007fb8b4007200 (allocated at /Users/tmcafee/sandbox/forestdb/src/filemgr.cc:233, size 4120) is not freed
total 2 objects

previous file:

../../forestdb_dump e2edb_main.0
DB header info:
BID: 133913 (0x20b19, byte offset: 548507648)
DB header length: 85 bytes
DB header revision number: 1012
HB+trie root BID: 133648, 128-byte subblock #0 (0x20000000020a10, byte offset: 547422208)
Seq B+tree root BID: 133648, 128-byte subblock #28 (0x3c000000020a10, byte offset: 547425792)
DB header BID of the last WAL flush: 133913 (0x20b19, byte offset: 548507648)

1. documents in the main index: 0 / in WAL: 0 (insert), 0 (remove)
2. live index nodes: 0 (0 bytes)
   Total document size: 0 bytes, (index: 0 bytes, WAL: 0 bytes)
3. KV stores: 1
   KV store name: default
4. documents in the main index: 0 / in WAL: 0 (insert), 0 (remove)
5. live index nodes: 0 (0 bytes)
   Total document size: 0 bytes
   Last sequence number: 0

Next file after compaction: e2edb_main.1