The idea is specifically about Windows.
Currently, thee's an effort to create an auto-updater, which would finally re-enable the functionality in "Check for Updates" dialog, doing that in efficient way, i.e. downloading and installing some small-sized patches.
While that seems reasonable, the problems implementing that seem too much.
We have another problem, that there's no easy way for user to choose installer UI language different from auto-detected by msiexec (one could play with "Region and Language" -> "Format", or use command line like "msiexec /i LibreOffice_X.Y.Z_Win_x64.msi ProductLanguage=1033", but that all is outside of user's scope).
So my proposal is this: could we simplify the auto-update function to simply checking if a newer applicable .MSI exists, and downloading it and launch? This would of course need to download ~300 MB MSIs instead of some smaller packages. But that would still improve things for wast majority of users who don't care much about that size downloaded; and even improve things for the rest who care (if not in download size, then in ease of update process - at least, that solution would be in no way worse than status quo). The auto-updater could take care of downloading and launching the MSI installer with UI language matching current user UI; which gives us all the code necessary for the small bootstrap application downloadable from the main site, which could do the same: search for latest applicable MSI on TDF site, download it, ask user for installation UI language, and launch MSI with that language. Having that bootstrap downloadable from main site as a download option (also requiring user to only download one small thing, and not bother with other downloads) would not exclude availability of MSI itself for those who need it to e.g. enterprise deployments using GPOs, etc.
Doing incremental updates as Markus has prototyped for Linux and Windows builds remains viable approach. This would perpetuate the status quo of monolithic packaging.
But I'm not opposed as it moves the delivery UX forward.
But it begs the question about stability and long term infrastructure support (how do we validate what is propagated out onto download mirrors that is then an automated install).
Some other download security concerns as laid out by Kendy in the see also bug 74934
And does this end efforts for Mozilla incrementals on Windows?