Could Airflow releases occur more frequently?

[karakanb]

This is something that I have been observing for quite some time and couldn't find the answer yet, so I thought "why not start a conversation around it?".

Here are the current release dates and the number of days between releases:

• Airflow 2.3.1 -> 2022-05-25 (25 days)
• Airflow 2.3.0 -> 2022-04-30 (26 days)
• Airflow 2.2.5 -> 2022-04-04 (41 days)
• Airflow 2.2.4 -> 2022-02-22 (63 days)
• Airflow 2.2.3 -> 2021-12-21 (36 days)
• Airflow 2.2.2 -> 2021-11-15 (17 days)
• Airflow 2.2.1 -> 2021-10-29 (18 days)
• Airflow 2.2.0 -> 2021-10-11 (23 days)

Looking at the days, it seems like on average there'll be a new release every ~31 days, regardless of patch version or minor version.

Number of patch versions:

• 2.0: 3 versions
• 2.1: 5 versions
• 2.2: 6 versions
• 2.3: 2 versions so far

Quite often I see there are bugs being reported, and promptly being fixed by the members of the community, or the maintainers. However, it takes quite some time to get these fixes landed on the official releases; which means in order to get the fixed versions in our environments we need to take care of finding a way to patch our images, or build custom Docker images on top of the master source code somehow.

The main point I wanted to discuss is: would it be possible for Airflow to have a more liberal release policy, with patch versions being released often? That'd mean:

• pretty much every new fix would land in a new patch version
• the difference between various patch versions would be minimal
• the upgrades between patch versions would be simpler and less risky from a sysadmin perspective
• time-to-fix in prod environments would be reduced dramatically
• any minor airflow version would likely have 10+ patch versions

Is this possible from an automation standpoint today? if yes, what would be the arguments against releasing often?