The only difference I could find between the two log files is this entry in the second one:
[2020-09-10T00:39:06.040 CEST DEBG ][ org.kde.kstars.ekos.capture] - setMeridianFlipStage: "MF_NONE"
When capture sets the meridian flip stage to "MF_NONE" directly after receiving a "FLIP_COMPLETED" this means that capture thinks that there is no capture sequence job left to continue - i.e. it assumes that all jobs are completed.
This explains why there is no re-alignment and no guiding startup, but it does not explain why focusing and capturing continues. Strange...