When you undelegate, your GRT does not come back. It enters a thawing period of 28 days, at the end of which you can withdraw it. During that period it earns nothing and you cannot move it, redelegate it, or change your mind.
That is the whole mechanic. Everything interesting about it is consequence.
Why a delay exists at all.
The delay is not a fee and nobody profits from it. It exists because the protocol’s security depends on stake being present when misbehaviour is discovered, and misbehaviour is discovered after the fact.
An indexer serves a wrong answer. Somebody notices, compares proofs, and opens a dispute. Arbitration takes time. If capital could leave the moment the wrong answer was served, every economic guarantee the protocol offers would be conditional on nobody noticing quickly enough.
The thaw is the window in which consequences can still attach to capital. That is also why the delay is measured in weeks rather than blocks: it has to outlast the process of catching someone.
What it costs you, concretely.
Suppose you delegate for three months and then leave. Your capital is committed for three months plus 28 days, and earns for three months.
Roughly speaking you have given up about a quarter of the return you thought you were getting, and no fee was charged, and no page told you. Rotate quarterly and you pay that quarterly. Leave the position alone for two years and the thaw rounds to nothing.
The rewards calculator puts real numbers on this. The pattern is worth internalising directly, though: the thaw punishes activity, and rewards leaving things alone. Any strategy involving frequent movement between indexers is fighting the mechanism rather than using it.
What Horizon changed, and what it did not.
Two changes, both improvements, neither touching the length of the wait.
The delegation tax is gone. Entering used to cost a percentage on deposit, burned. It is now 0%, removed from the protocol entirely. Note that the tokenomics page has not caught up and still describes the old charge, which is recorded on the parameters page.
You can queue more than one undelegation at a time. Previously a delegator was restricted to a single active undelegation, so exiting several positions meant serialising them and waiting through the thaw repeatedly. Under Horizon they run in parallel.
Practical consequences.
- Decide the holding period before you delegate, not after. The thaw makes an exit an event rather than a click, and the maths only works if the intended period is long relative to it.
- Do not delegate GRT you might need. This is the plain version. If a four-week delay to access would be a problem, this is the wrong place for that capital.
- Start the undelegation as soon as you have decided. The clock starts when you submit, not when you finish deciding. There is no benefit to waiting.
- A worsening indexer is a slow problem. If an operator raises their cut or stops allocating, you cannot react quickly. This is precisely why the selection criteria in the previous lesson lean so hard on durability rather than on rate.
Before reading on: why is the thaw measured in weeks rather than in blocks or epochs?
Because it has to outlast a human process, not a computational one.
The thaw exists so that stake is still reachable when misbehaviour is found and arbitrated. Finding it involves somebody noticing a mismatch, raising a dispute, and arbitrators reaching a decision. That is a process measured in days and weeks regardless of how fast the chain runs.
Denominating the period in blocks would tie a human deadline to a machine clock, and would need changing every time block times changed. Weeks are the honest unit for the thing being measured.
During the thawing period, your GRT:
It is inert. That is the entire cost, and because it is not charged as a fee it never appears in an advertised return. You have to compute it yourself.
Why does the thawing period exist?
Nobody profits from the delay. Misbehaviour is discovered after it happens, and disputes take time to resolve. The thaw is the window in which consequences can still attach to capital, which is why it is measured in weeks rather than blocks.
Which strategy does the thawing period penalise most?
Every rotation costs a thaw. Quarterly rotation surrenders a meaningful share of the return to periods where capital is committed and idle. The mechanism rewards leaving things alone, so a strategy built on movement is fighting it.
0 of 3 answered