This was just a proof of concept test that I added when working on #5981; however it's not very useful in travis (it needs multiple bitcoind binaries to do anything useful), and it's unbelievably slow, and so it was just serving as a reference for other test writers interested in local tests comparing multiple binaries.
However lack of use also leads to inadvertent breakage (as pointed out here: #6971 (comment)), so I think it's probably time to remove.
lack of use
But didn't @luke-jr use it in #6971 (comment)?
serving as a reference
Is there a "new and better" way to test multiple bins?
@MarcoFalke The framework for testing multiple binaries is still in place. The methods I used in this pull (especially using invalidateblock/reconsiderblock over and over) should probably be done a better way somehow, but I haven't thought about how I'd do it again if I were going to reimplement.
As for the meat of this test, it is just implementing the script tests that are performed by the unit tests already anyway (except over the p2p network), so we're not losing any test coverage functionality that I can think of.
The breakage in #6971 (comment) is related to the test framework more than script_test.py AFAICT...
ACK, if no one complains.
ACK.
ACK
Removing unmaintained and dead code is good practice. Anyone that wants to 'resurrect' the code can always get it from git history.
Speaking of which - I'd really like to re-enable the rpcbind.py test without the IPv6 part until the a new stable libevent is in common use. With libevent's release schedule it may take years otherwise :-)
utACK
ACK.