Philip,
There's one thing here I should have said more clearly, because it changes the scale considerably.
CONTRIBUTION IS PER ROUND, NOT PER YEAR
A "round" is the time between two pool blocks. When a block is found, everyone's accumulated contribution is paid out and the counters reset. So a worker never carries a year of shares — it carries whatever it built up since the last block. An S21 that ran for a year and then broke has nothing at risk except its share of the current round.
TWO WAYS TO SWAP A MACHINE — BOTH WORK
Option A — same worker name (what you did with the T21 on the Apollo's name)
The new machine simply inherits the old worker's round contribution. The pool sees one worker whose hashrate went UP. The hashrate check compares a worker to its own average, so a bigger machine on the same name passes instantly — no penalty, nothing lost, and the old contribution keeps building under the new hardware. For an UPGRADE this is the best possible move.
The one case to avoid is the reverse: a SMALLER machine on the old name. Then the pool sees that worker drop (33 → 10 TH) and, after 5 hours below 85% of its own average, starts writing it down — because it can't tell a downgrade from a rental leaving.
Option B — new worker name (EPIC controller, unique ID, whatever it forces)
The new machine starts its own clean record immediately. The old worker's round contribution stays on the books and is paid at the next block, less the offline ramp for however long it's been unplugged:
Time from unplug to the next pool block Old unit's round contribution
--------------------------------------- -----------------------------
under 4 hours paid in full
4 – 8 hours paid, minus a few %
8 – 16 hours paid, minus an increasing share
16 hours or more not paid
SO WITH EPIC CONTROLLERS
You're fine. Option B is always available, and the exposure is one round's share of the old unit, only if the round runs more than a few hours past the swap. If the old unit can still hash at all, leave it plugged in beside the new one until the next block, then pull it clean.
And on your actual day: the Apollo's contribution was written down for the hours it was dark, in proportion.
Nothing the T21 earned was touched under either name.
Hope that makes sense.
Okay its flawed but not much to fix it.
If the worker apollo was going bad and ran 70-30 for 1 month at 10th then needs replacement . How to replace it and not sacrifice the shares.
The internet sucks at times when we try to talk with one and other.
My goal was to replace the failing 8-10th unit with a larger unit.
Which I did and still lost the points/hash/potential rewards on the bad unit.
So if a guy is running a s21 50-50 for a year it breaks and he switches in a s21xp he loses a year's hash unless he matches the workers id exactly.
First unit is 200th
Replacement unit is 270th must match the id's. Or lose the hash.
The issue with the t21 is the epic controller forces a unique id. Which will not match the apollo unless I had to match worker id's not to lose hash.
So if I use any epic controllers on your pool I have a risk i can't fix with a second miner using an epic controller.