but I moved zinc off binary analysis format - they were using JVM serialization, which was waaaay slower than the text format. Any idea why they’re moving back?