Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.95%. Comparing base (854913c) to head (9a9815a).
⚠️ Report is 5 commits behind head on master.

@@           Coverage Diff           @@
##           master    #3311   +/-   ##
=======================================
  Coverage   99.95%   99.95%           
=======================================
  Files         380      380           
  Lines       53912    53916    +4     
=======================================
+ Hits        53890    53894    +4     
  Misses         22       22           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

IT passed as well, a minor change with big impact (on selected few files ;) )

I am taking this back for now. While one case (above) seems consistent, I re-did some measurements, and I am getting some inconclusive results ... it could be that the mn416-branch that I used missed some updates to the dependency_tools?

I've verified that the 'speedup' of the DA on master was caused because the references were not resolved (turned up as unknown, so DA aborted early). Resolving the symbol fixes the 'speedup' (which marked the loop as not parallelisable, even though it is), it is then slow again with the correct result, and this patch fixes it:

master:
ukca_um_strat_photol_mod.F90	2	Execution time loop i:	14.372405	True	
ukca_um_strat_photol_mod.F90	3	Execution time loop j:	10.499314	True	
ukca_aerod.F90	                1	Execution time loop j:	35.180508	False	

3310:
ukca_um_strat_photol_mod.F90	2	Execution time loop i:	0.155515	True	
ukca_um_strat_photol_mod.F90	3	Execution time loop j:	0.132040	True	
ukca_aerod.F90	                1	Execution time loop j:	0.695516	False	

Looks like no changes, so IT do not run again. Ready to go. Note that there are still some outliers when running the DA on ukca. I've added some additional potential improvements to #3183 (#3183 (comment))