The first two warnings about the non final TOO_SMALL fields have been discussed on the list (see http://markmail.org/message/tnxlul6qfjzs3mr3). Phil proposed to comment this in the release notes.
The next four floating point equality warnings and the last four internal representation exposed are intentional. They are identified in a commented findbugs exclusion filter in the subversion tree (file findbugs-exclude-filter.xml added about three weeks ago). This filter is active when findbugs report is created from maven2 (see plugin configuration in pom.xml).
The null pointer possible dereference was also discussed here (see http://markmail.org/message/33jpeig4qjmvtvy7). I forgot to check after Phil answered. I'll do it.
I'll fix the static inner class in Frequency if possible without breaking compatibility.
I understand the concerns about synchronisation, but would prefer someone else check them also.
As a summary, only the first two warnings should remain at the end.
Thanks for the report.