If any subdirectory of the home directory isn't readable, Solr fails as indicated above.
Yeah ... that part really wasn't clear from the initial bug report.
The issue here is that in core-discover mode, solr really can't tell if a non-readable directory is a sign of a problem or not. The flip side of ignoring non-readable directories is that it would mean solr might happily startup w/o some core you are expecting to be there if the permissions are set wrong.
...Despite this otherwise being an otherwise healthy Solr home directory, Solr fails because it can't read lost+found.
The specific situation you show is a great example of why i would argue that the presense of a non-readable directory like lost+found is in fact a serious problem and you might not want thi solr node to startup because of it. what if lost+found contains your entire collection?
regardless of wether we think solr should ignore non-readbale files, we should at least generate a better error message then NullPointerException