A slot round is a sequence with a beginning, a stake and a result, and everything a studio builds between those three points can be lost in the second between them. The connection drops, the tab is closed, the handset restarts, the game server is redeployed — and a stake has left the player’s balance while the result that stake bought has not yet reached the screen.
Define the interruption policy before implementation. The commissioning brief should explain how a paid round completes or voids, how a free-spin sequence resumes and how progressive contributions are recovered. Later changes remain possible, but may require engineering, data migration and market-specific review of the changed build.
This article sets out what the published standards actually require, what state has to survive an interruption, and the eight questions that belong in the brief before a line of the client is written.

What the British standards say, and what they do not
Great Britain’s published standards give a useful starting point for an interrupted-round recovery brief. Read both the requirements and their product applicability before translating them into implementation decisions.
Remote technical standard 10 carries the aim “to ensure that customers are treated fairly in the event of interrupted play or betting and that they are aware of how they will be treated if interruptions occur”. Requirement 10A says operators must take all reasonable steps to ensure their policies for instigating or dealing with service interruptions “are fair and do not systematically disadvantage customers”. Its implementation guidance then becomes concrete: where an interruption occurs after the operator receives notification of the customer’s gamble and the customer can have no further influence on the outcome, the results of the gamble should stand; where an interruption to a single-participant single-stage event occurs before an outcome has been generated, the deducted stake should be returned; and for stateful games — games with multiple stages or decision points — all reasonable steps should be taken to restore the game to its last known state so the customer can complete it. Progressive jackpot values should be restored to their pre-failure state.
Requirement 10B moves from policy to capability. Systems must be capable of recovering from failures that cause interruptions, including “where appropriate” the capability to void gambles with or without manual intervention, to suspend betting markets, and to take all reasonable steps to retain enough information to restore events to their pre-failure state. The guidance lists what that information is: the state of a deck of cards and hands dealt, the number of tokens collected, other predetermined information such as maps or prize layouts, the value of any progressive jackpots, the state of any gambles, and the bets placed or offered. Requirement 10C adds the disclosure duty: operators must make information about their interruption policies available.
The applicability line needs to be read alongside the Commission’s product applicability annex. The annex marks RTS 10 as applicable to casino, bingo and peer-to-peer betting. Its reference to gaming is not limited to peer-to-peer games: a single-player casino slot cannot be treated as exempt on that basis. The RTS 10 guidance distinguishes completed outcomes, undetermined stakes and stateful games; the recovery plan must match the product and failure state.
That general standard matters here. RTS 5 requires all reasonable steps to ensure gambles are “accepted, processed and settled in accordance with the operators’ published terms and rules”, and adds that where unexpected system flaws, faults or errors affect the customer, steps must be taken as soon as practicable to remedy the problem and treat the customer fairly. In other words: whatever the interruption policy is, it has to be in the rules and the system has to implement it.
An older standard puts the single-player rule plainly
Gibraltar’s published remote technical and operating standards state the single-player case directly, and provide a practical design benchmark. Section 7.7 of the 2012 RTOS guidance asks licence holders, where possible, to provide a mechanism in single-player games “for an incomplete game to be completed before a customer is able to participate in any other game”, and names the causes it is written for: loss of communications between the licence holder and the end customer device, a licence holder restart, a game disabled by the licence holder, an end customer device restart, and abnormal termination of the gambling application on the device. It adds that gambles associated with a partially complete game that can be continued should be held by the licence holder and be apparent to the customer.
The same document’s record-keeping guidelines describe the data that makes this possible: gambling event and transaction history should include, as appropriate, the session’s wagers time-stamped, the event result, the display or symbols determining the final outcome, the choices made by the customer, the account balance at the start and end of the session, the event status — complete, in progress — and the credits or bonus features awarded in play.
That list is the practical definition of restorable state, and it is a fair test of any studio’s build. If sufficient authoritative recovery records were not retained, reconstructing the interrupted state may be impossible. Those records can span the session, outcome, wallet and event logs; no single session-status field is the only valid design. These standards are published as a versioned guidance document rather than a substitute for a current market’s licence conditions, so a buyer should treat them as a design benchmark and confirm the specific duty against the market being entered.
The state that actually has to survive
Strip away the language and an interrupted slot has to answer five questions about what it kept.
Did the stake leave the balance, and is that recorded as a settled transaction or an open one? This is the hinge. A stake deducted against a round that never resolved is the one state a player will always notice, and it is the one an operator’s support queue will always hear about.
Was an outcome generated before the interruption? If it was, the result exists and the question is only how it is presented; if it was not, there is nothing to present and the stake goes back. The British guidance draws that line for single-stage events, and it is the right line for a slot: the outcome is a fact once the RNG has produced it.
What was the feature state at the moment of the interruption? Free spins remaining, a held multiplier, a stack of collected tokens, a decision point awaiting input, the reels held on a line. Incorrect recovery may change entitlement or misrepresent progress unless authoritative state and settlement are preserved. Replaying presentation alone does not necessarily change a predetermined outcome or payout.
What did the jackpot pool look like? A progressive contribution taken from a stake that never produced a result, a jackpot triggered but not yet awarded, a pool that must return to its pre-failure value rather than to a rounded one. Jackpot state is the part of recovery that cannot be reconstructed from a client log.
Can the customer see any of it? A restored round is only fair if it is visible — the stake held, the round in progress, the feature at three spins of ten. Silence reads as a lost stake even when the balance is correct.
Recovery is a server property, not a client trick
A common brief asks the front end to “remember” a round in local storage and replay the animation on reload. That is not recovery. It is a presentation of a state the client guessed at, and it fails the moment the player changes device, the browser is cleared, or the tab is closed at the wrong microsecond.
The recovery design should keep authoritative round state on the game server or RGS, linked to a round identifier the client can query on reconnect. The client asks for that state and presents the response. The brief should specify when an unresolved round blocks further play, matching the platform’s behavior and the applicable market requirements. On the wallet side, define how each incomplete round completes or voids and make settlement idempotent so a retried request cannot debit or credit twice.
This is where commissioned and licensed games differ commercially, not technically. A studio that builds on an existing engine inherits whatever session and round model the platform already enforces, and the brief becomes a set of questions about the platform. Hiring a studio to finish an existing engine game covers how to scope that inheritance. A studio writing its own server owns the state model outright, and the recovery behaviour is then part of what it is being paid to design.
The boundary with latency, and why the two get confused
Time-critical events are a separate duty and a separate page of the same standards. RTS 4 asks operators, where speed of interaction has a significant effect on the player’s chance of winning, to assess the risk and demonstrate reasonable steps to reduce it — estimating network latency and telling the customer about any disadvantage, handicapping by measured performance, or treating wins that arrive within a window as simultaneous. It also requires the customer to be told that they may be at a disadvantage because of slower networks or devices.
A slot’s spin button is not a time-critical event in that sense: the round resolves from a generated outcome, not from who pressed faster. The confusion is expensive because the two duties lead to different work. Latency asks for disclosure and, where it genuinely matters, handicapping; interruption asks for stored state and a completion path. A brief that mixes them produces a game that shows a connection-quality badge and still loses a free-spin sequence.
The same discipline applies to jackpot design. A jackpot’s rules have to describe how prizes are determined and awarded “including what happens when two or more players simultaneously trigger the same jackpot, or appear to simultaneously trigger the jackpot, for example due to network latency issues”. The progressive jackpot commissioning analysis covers funding and settlement; the interruption question is what the pool does across the restart, and it belongs in the same document.
Eight questions for the brief
Each of these has a testable answer, and each answer changes the build.
- What is the round identifier, and can the client ask the server for an open round by it after a reconnect?
- Is a stake an open transaction until the round resolves, and what is the timeout that closes or voids it?
- If an outcome was generated, how is it presented after reconnection — and is the presentation identical to the interrupted one?
- Which feature states are persisted, and at what granularity: spins remaining, multiplier, collected tokens, pending choice?
- What happens to a progressive contribution from an unresolved round, and how is the pool restored to its pre-failure value?
- Is settlement idempotent, and what does the server do with a duplicate completion request?
- Can a new stake be blocked while an unresolved round exists, and what does the player see when it is?
- What is written into the customer-facing rules about all of the above?
The last question is not decoration. The disclosure duty is real in Great Britain, and the rules page can be central to resolving a dispute. The studio commissioning checks cover the wider handover set, and the math-to-client parity evidence covers the sibling problem of proving that what was designed is what shipped.
The evidence to ask for before launch
Recovery is testable, and the test is cheap compared with a launch defect.
Take a build, deduct a stake, interrupt the connection mid-spin and mid-feature, then reconnect. The round should resume or void, the balance should reconcile either way, the feature should be at the stage it was at, and the session record should still say in progress or complete rather than nothing at all. Repeat with the device restart, with the tab closed between the stake and the result, and with a duplicate completion request replayed at the server. Then read the transaction history and check that it tells the story: stake, outcome, feature credits, final balance, status.
Two evidence artefacts are worth having in the handover pack. The first is a state table: every round state the game can be in, what persists it, and what the recovery path is. The second is a short test transcript per state, tied to the build identity. Neither is a certification document. Both are what an audit, an operator’s integration team, or a market’s technical dossier will ask for, and both are much cheaper to produce during development than to reconstruct afterwards.
Backup and restore verification covers the platform-level version of the same discipline, and the integration tests an aggregator runs is where a recovery gap is most likely to surface first, because an aggregator’s acceptance pass includes exactly these reconnect cases.
Who owns which half
Split the responsibility explicitly, because it is the part of this topic that goes wrong commercially.
The studio owns the game-side state model, the round identifier, the resume path and the feature persistence. The operator or platform owns the wallet transaction semantics, the timeout that voids an abandoned round, the rules text the player reads, and the decision about whether an incomplete round blocks further play. The aggregator, where there is one, owns the session handshake and the mapping between its round reference and the studio’s.
A brief that assigns all of it to “the game” produces a game that does what it can and a launch that discovers the wallet half in production. Written as three columns with names against them, it produces a testable handover.
If you are scoping that handover and want the round model, the state table and the evidence plan settled before engineering starts, request a project quote. The commissioning guide covers how the rest of the brief is assembled.
Questions buyers ask
Is an interrupted slot round a regulatory requirement or a design choice?
Both. The Commission’s product applicability annex includes casino under RTS 10, whose guidance addresses recovery of stateful games. RTS 5 also requires settlement in line with the operator’s published terms. Gibraltar’s guidance addresses incomplete single-player games, with a qualification that the completion mechanism should be provided where possible. A studio should confirm each market’s requirements and turn them into a documented, testable recovery design.
What happens to the stake if the connection drops before the result appears?
It depends on whether an outcome was generated. The cited guidance distinguishes returning a stake before an outcome exists from preserving a result already determined. The system needs linked round, outcome and wallet records sufficient to identify the failure state and reconcile settlement. This does not require every platform to keep the stake debit itself open until the round completes.
Can the front end store the round and replay it after a reload?
It can store presentation hints, but it cannot be the record of the round. A client-side copy disappears with the browser cache and differs between devices, so a player who reconnects elsewhere loses the round or sees a different one. The restorable state belongs in the game server or RGS session, keyed by a round identifier, with the client asking for the open round on reconnect.
How much feature state has to persist?
Enough that the customer can complete the round they paid for, at the stage they reached. In practice that means the spins remaining in a free-spin sequence, any accumulated multiplier or token count, a pending choice in a decision-based feature, and the held symbols where the game holds them. The recovery plan should preserve authoritative state and settlement and explain the player-facing behavior. A presentation replay with the same recorded outcome is different from changing the player’s entitlement.
Does a progressive jackpot contribution survive a restart?
It has to be reconstructable, and the pre-failure value has to be restored rather than recalculated. The published interrupted-gambling guidance names the value of any progressive jackpots as information the system should retain and restore, and jackpot rules separately have to explain how prizes are awarded when two players appear to trigger the same jackpot simultaneously, including through network latency. Both belong in the jackpot design document before the game is built.









































