Add test reproducing hunk locking problem [do not merge]

[mtsgrd]

trafficstars

[daniilS]

@mtsgrd #4136 is another case that triggered the same error, but without changing line numbers - not sure if that would need its own regression test, or if it's also covered by this one.

[mtsgrd]

@daniilS the two are definitely related, in that they cause(d) the same kind of panic. This one is preventable though, it's a matter of locking the change to the branch it depends on.

[krlvi]

@mtsgrd does the new implementation cover this case now?