[QGIS-Developer] LTR 3.10: Fixing issues when content cache is used in blocking mode

[René-Luc Dhont]

Hi devs,

I need your point of view about fixing issues when content cache is used 
in blocking mode.

A QGIS Server user has identified an issue when content cache is used in 
blocking mode. In some cases, the bug is timed out QGIS Server.
This issue is due to a loop to wait for content task cache begun. In 
some cases, the loop starts even if the task has already begun, and it 
is never stopped.
This issue has been fixed in master and version 3.12, by fixing the task 
manager waitforfinished (PR 32838 [1], a 6 mounths old PR) and removing 
the loop when content cache is used in blocking mode (part of PR 35158 [2]).
These pull requests has not been backported to LTR, the 3.10 release.

To fix, the issue in LTR, I have manually backported the 2 Pull Requests:
* Backport Fix Task manager waitforfinished, PR 36651 [3]
* Backport Fix issues when content cache is used in blocking mode, PR 
36624 [4]
And requested Nyall Danwson reviews.

After Nyall reviews, I have two options to fix the issue:
1. Fix Task manager waitforfinished (merge PR 36651 [3]) and removing 
the loop (removing commit c184270 [5] from PR 36624 [4] then merge it)
2. Keep and fix the loop (merge PR 36624 [4] and close PR 36651 [3])

So I need a collegial decision about the least bad way to fix the issue:
1. Merge a risky backport and remove the loop
2. Keep and fix the loop

Thanks for your feedbacks.

René-Luc

[1] https://github.com/qgis/QGIS/pull/32838
[2] https://github.com/qgis/QGIS/pull/35158
[3] https://github.com/qgis/QGIS/pull/36651
[4] https://github.com/qgis/QGIS/pull/36624
[5] 
https://github.com/qgis/QGIS/pull/36624/commits/c184270ebe2bb51164fdcd13b7ccd8ebc42e71a0