OpenZFS 2.3.0-rc1 tagged: It's the first full development cycle the author was involved in.

• Better userspace debugging: Extended zdb output to include thread information and stack traces for better debugging of userspace issues when recovering data from damaged pools. Also set thread names for easier identification during process inspection.
• Better task queue stats: Replaced the old task queue stats with more useful kstat nodes, showing instantaneous and accumulated values for internal task queue operational state. The original PR also included rolling averages but was removed due to overhead.
• The endless kernel update treadmill: Keeps OpenZFS updated with new Linux kernels (up to 6.12-rc1) and removes support for old ones. Experimental support is available in point releases. Detects the kernel version during compilation and aborts if it's newer than the maximal supported version.
• Linux IO submission rewrite: Took a year to solve properly. Rewrote the IO submission layer to be smarter and more efficient, fixing an alignment issue and improving write performance by 10% on heavy sync writes and 150% on async writes. Also enables "zero copy" facilities.
• Credit where it’s due: Maintains a list of contributors in the AUTHORS file and updated it by mining the git history.
• JSON everywhere: Many zpool and zfs subcommands now take a -j switch to output JSON. Allows for easier parsing and integration with other tools.