Snapshot builds for 0.0.x versions are generally built after the integration of new bug fixes, having passed the basic test harness. They can be considered generally stable mid-releases in this sense.
It is not uncommon however, for bug fixes to have unintended consequences, which is why official releases are the result of more thoroughly tested code. Official releases have also passed the full parser fuzz-testing, and comparative regression tests(1). Since these tests are intensive and take several hours to run, they are not performed for every snapshot release.
(1) Comparative regression testing is a custom testing harness designed to generate random patterns of compilable and executable code with known return values. Third-party scripting engines (BeanShell, Groovy) are used to execute the code for comparisons