Theory on Issues with Lyra Coins

First of all, I'm new to mining in general, and most of this is speculation based on comparing results here to elsewhere (and the volume of complaints about payments for these two cryptos - barring some of the "nicehash" noise of people coming from far larger pools, and expecting quicker results).

I believe there are probably 2 issues server side - and I think they are linked:

1) VarDiff - This seems to want each miner to have a super fast turn around, e.g. it's happy with them being returned within a couple of seconds (if that).  Other pools (for VTC) seem to set my difficulty around 64 (with sometimes dropping to 32, and then fairly soon going back up again) for my single GTX1080 - here it always hovers around 18.  This means on other pools I sometimes get shares that take 30 seconds or more - and others have a bunch within a couple of seconds.

E.g. this on P2Pool:

For us here, I think this is a "sledgehammer to crack a nut" scenario - using higher powered devices to solve simpler problems - people get a high number of shares - but because they are the small fry problems they aren't worth as much as their GPU solving potentially much harder stuff in a similar amount of time.

2) Allocation of jobs - this comes from the Excavator issue that is in another thread and that it seems like only here that CCMiner reports "enabled stratum stale jobs workaround", which from my investigation in another thread is that the server is reporting that the job doesn't exist - i.e. it's already been solved by someone else.  You can see this happens a lot in Excavator, and I believe CCMiner is just masking it by verifying that the job hasn't gone stale before reporting the share.  Given this I think the server doesn't allocate jobs efficiently, and will give out jobs to multiple workers, and whoever solves it first gets the share, and CCMiner will ditch the job before attempting to send the result - i.e wasting compute resource.

I would say these possibly go hand in hand, VarDiff would need to allow for more time between shares, however, the allocation of jobs needs to be so that multiple workers aren't working on the same jobs within that window.  It's possibly why the VarDiff target is so low.

I could of course be completely off base here, but thought I'd throw it out there :)