Automatic sanity-testing: PASSED, see http://jenkins.bluematt.me/pull-tester/p4803_b4ad217d8f6e9beb5268e27ba77a105a20a7f2f4/ for binaries and test log.
This test script verifies pulls every time they are updated. It, however, dies sometimes and fails to test properly. If you are waiting on a test, please check timestamps to verify that the test.log is moving at http://jenkins.bluematt.me/pull-tester/current/
Contact BlueMatt on freenode if something looks broken.

Looks good to me but you should very carefully test this. The datadir determination is extremely brittle code.

Many times, changes such as this have broken something like 'Set the datadir in bitcoin.conf' or Set the network in bitcoin.conf. Also be sure to test with both the GUI and bitcoind as they have different initialization sequences.

Ok, I though this could be controversial and that's why I separated it. I'll test this within #4804
Closing.

The datadir stuff is really overdue for a cleaner rewrite.

The caching makes little sense here in the first place. It should simply compute the directory (once!) and store it when the datadir (and the network, for the specific dir) is set, instead of computing it when it is requested and then caching. But this means having to properly define the initialization order :-)

I've postponed this myself as this requires a comprehensive test plan first (what are people actually using?). And well, it works* and there are zillions of more serious issues.

• There are some issues such as bitcoind --help wants write access to file system #3639 that could be solved here

Btw, I know @gmaxwell is interested in removing worthless caching lately...