(topic deleted by author)
If the “inconsistency” you’re talking about is on-again-off-again studying, with a big backlog of overdue cards – I don’t think FSRS is doing quite what you think it’s doing.
When you study an overdue card, FSRS already expects that you have a lower chance of remembering it, so if you get it wrong, there’s no surprise. That doesn’t skew your parameters in any particular way, it just teaches FSRS how your memory decays. I don’t see anything in your parameters that suggests an issue [thanks for including them!].
If you’ve got a heavy workload – and it’s not due to having a backlog right now (or having recently dug out of a backlog by studying more cards each day than usual) – I think it’s more likely because your retention outcomes are still lower than your Desired Retention [DR].