New Marlin 2.0 G29 behaviour

I just upgraded my Mini (1) to Marlin 2.0.0 using the Lulzbut Marlin git repo devel branch, which compiles nearly perfect (couple of warnings) and flashes well. There was one awkward change in G29 behaviour though:

My preprint custom g-gcode amongst other things puts everything in absolute mode (G90) homes all axis and heats to 170C to prepare for leveling with G29. Just before G29 I slowly retract some filament using G1 Z150 E-20 F25 and after this finishes start the G29, just like the original Lulzbot start g-code.

In the new firmware, the G29 nulls the E-20 of the previous move, the Z150 seems to be honoured but it may very well be eaten by the G29 as well, but there is no filament retract. I have to put an M400 after the G1 Z150 E-20 to succesfully retract 20mm of filament.

Does anybody know if this new behaviour is by-design?

Also, my mini now suffers from this issue:
It’s not bad enough to downgrade. Just a heads-up for those contemplating the upgrade.

Raised an issue for the encountered behaviour:

The retract before probing bug has been fixed, as well as the FW_RETRACT micro pauses.