For completeness' sake... I've tried a dozen runs at this already, looking for tuning parameters to tweak. It always starts well and then degrades rapidly before finally hanging.
MDBM: version 4.11.1
Date: Tue Dec 16 20:34:36 2014
CPU: 16 * Intel(R) Xeon(R) CPU E5-4650 0 @ 2.70GHz
CPUCache: 20480 KB
Keys: 16 bytes each
Values: 2000 bytes each (1000 bytes after compression)
Entries: 76800000
RawSize: 147656.2 MB (estimated)
FileSize: 74414.1 MB (estimated)
------------------------------------------------
2014/12/16-20:34:38 ... thread 0: (200000,200000) ops and (128923.6,128923.6) ops/second in (1.551306,1.551306) seconds
2014/12/16-20:34:39 ... thread 0: (200000,400000) ops and (110571.0,119044.1) ops/second in (1.808792,3.360098) seconds
2014/12/16-20:34:42 ... thread 0: (200000,600000) ops and (87921.9,106480.3) ops/second in (2.274746,5.634844) seconds
2014/12/16-20:34:44 ... thread 0: (200000,800000) ops and (107459.0,106723.3) ops/second in (1.861175,7.496019) seconds
2014/12/16-20:34:46 ... thread 0: (200000,1000000) ops and (90923.1,103138.7) ops/second in (2.199660,9.695679) seconds
2014/12/16-20:34:50 ... thread 0: (200000,1200000) ops and (51852.2,88542.6) ops/second in (3.857118,13.552797) seconds
2014/12/16-20:34:53 ... thread 0: (200000,1400000) ops and (57516.7,82207.6) ops/second in (3.477253,17.030050) seconds
...
by the end
2014/12/16-21:18:19 ... thread 0: (200000,22400000) ops and (858.0,8541.6) ops/second in (233.094661,2622.447456) seconds
2014/12/16-21:19:19 ... thread 0: (200000,22600000) ops and (3321.7,8424.5) ops/second in (60.210362,2682.657818) seconds
2014/12/16-21:20:28 ... thread 0: (200000,22800000) ops and (2880.3,8284.6) ops/second in (69.437524,2752.095342) seconds
2014/12/16-21:21:31 ... thread 0: (200000,23000000) ops and (3203.6,8171.9) ops/second in (62.429181,2814.524523) seconds
then a stream of these start showing up in dmesg
Dec 16 21:14:09 localhost kernel: XFS: possible memory allocation deadlock in kmem_alloc (mode:0x250)
Dec 16 21:14:30 localhost kernel: XFS: possible memory allocation deadlock in kmem_alloc (mode:0x250)
Dec 16 21:14:32 localhost kernel: XFS: possible memory allocation deadlock in kmem_alloc (mode:0x250)
Dec 16 21:14:34 localhost kernel: XFS: possible memory allocation deadlock in kmem_alloc (mode:0x250)
Dec 16 21:14:36 localhost kernel: XFS: possible memory allocation deadlock in kmem_alloc (mode:0x250)
and on and on until I kill the job.
Dec 16 22:19:01 localhost kernel: XFS: possible memory allocation deadlock in kmem_alloc (mode:0x250)
Dec 16 22:19:03 localhost kernel: XFS: possible memory allocation deadlock in kmem_alloc (mode:0x250)
Dec 16 22:19:05 localhost kernel: XFS: possible memory allocation deadlock in kmem_alloc (mode:0x250)
Dec 16 22:19:07 localhost kernel: XFS: possible memory allocation deadlock in kmem_alloc (mode:0x250)
Dec 16 22:19:09 localhost kernel: XFS: possible memory allocation deadlock in kmem_alloc (mode:0x250)
Dec 16 22:19:11 localhost kernel: XFS: possible memory allocation deadlock in kmem_alloc (mode:0x250)
Dec 16 22:19:13 localhost kernel: XFS: possible memory allocation deadlock in kmem_alloc (mode:0x250)
Dec 16 22:19:15 localhost kernel: XFS: possible memory allocation deadlock in kmem_alloc (mode:0x250)
Dec 16 22:19:17 localhost kernel: XFS: possible memory allocation deadlock in kmem_alloc (mode:0x250)
Dec 16 22:19:19 localhost kernel: XFS: possible memory allocation deadlock in kmem_alloc (mode:0x250)
Dec 16 22:19:21 localhost kernel: XFS: possible memory allocation deadlock in kmem_alloc (mode:0x250)
Dec 16 22:19:23 localhost kernel: XFS: possible memory allocation deadlock in kmem_alloc (mode:0x250)
Dec 17 00:14:16 localhost systemd: Got automount request for /proc/sys/fs/binfmt_misc, triggered by 1372 (vmtoolsd)
LMDB does this load in 8 minutes at an effective data rate of ~283MB/sec. Pretty much as fast as the hardware will stream (peak throughput of 300MB/sec in this VM). I don't see how you can possibly use MDBM as an unreliable data store if it could potentially take hours to reload the data.