Connecting independent networks has been attempted repeatedly through different approaches, and the outcomes differ substantially.
Atomic swaps
Direct exchange between chains using hash-locked contracts.
Which requires no trusted party and requires both parties to be online.
The coordination requirement limited practical adoption despite the elegant design.
Custodial wrapping
A custodian holds an asset and issues a representation elsewhere.
Which is simple and places full trust in the custodian.
The largest wrapped assets by value have used this model.
Validator-set bridges
An external group attesting to events on one chain for another.
Which is where most of the largest losses occurred.
Security equals the cost of compromising a threshold of signers.
Light client verification
Verifying another chain's consensus directly.
Which inherits that chain's security rather than adding an assumption.
Implementation cost and per-chain-pair effort have limited deployment.
Shared security frameworks
Chains within a common framework connected at the protocol level.
Which removes the external bridge entirely between participating chains.
It requires chains to be built within the framework, which is a substantial constraint.
Zero-knowledge approaches
Proving state transitions succinctly for verification elsewhere.
Which is the direction current work is taking.
Proving cost and latency are the practical constraints being addressed.
Liquidity networks
Pools on both sides with swaps rather than locking and minting.
Which changes the failure mode from custody compromise to liquidity shortage.
What the record shows
Approaches trusting an external party have failed repeatedly; approaches verifying cryptographically have been harder to build and have not failed in the same way.
This is technical description rather than a recommendation about any system.
Message passing beyond assets
General cross-chain calls allowing contracts to interact across networks.
Which is more useful and carries the same attestation question.
Failure handling for messages that succeed on one side and fail on the other is a specific design problem.
Standardisation efforts
Common interfaces so applications are not locked to one bridge.
Which is progressing and has not converged.
Applications supporting multiple protocols hedge against any single failure.
Value at risk
Total value held in cross-chain infrastructure is publicly measurable.
Which allows the aggregate exposure to be assessed.
It remains substantial and is concentrated in a small number of systems.
Insurance and backstops
Some protocols maintain reserves against bridge failure.
Which has funded partial reimbursement in past incidents.
The practical guidance
Prefer natively issued assets, minimise bridged balances, and check the trust arrangement before moving anything substantial.
Canonical versus third-party bridges
Official bridges operated by a network's own team differ from independent ones.
Which have different trust assumptions and different asset representations.
Multiple representations of the same asset from different bridges are not interchangeable.
Withdrawal delays
Optimistic designs impose challenge windows measured in days.
Which liquidity providers bridge for a fee.
Using those services substitutes provider risk for waiting.
Monitoring
Automated systems watching for anomalous bridge activity.
Which has caught incidents in progress.
Detection speed has determined loss size in several cases.
User interface abstraction
Applications hiding cross-chain complexity from users.
Which improves usability and obscures the trust assumptions being accepted.
The recurring conclusion
Where a system relies on an external group to attest, the security is the cost of compromising that group.
The conclusion the record supports
Every approach that trusts an external group to attest has been compromised, repeatedly, at scale.
Approaches verifying the source chain cryptographically have not failed in the same way, and are harder and more expensive to build, which is why they remain the minority.
Practical guidance
Prefer natively issued assets, minimise bridged balances, use canonical bridges where available, and check the risk trackers first.
Which reduces exposure to the category with the worst historical record in the field.
This is technical description rather than a recommendation about any specific system.
The measurable record
Incident databases catalogue every significant cross-chain failure with the amount lost, the mechanism and the response.
Reading through that catalogue takes an hour and produces a clearer sense of the risk than any amount of protocol documentation.
The clustering of causes is the finding: a handful of failure modes account for nearly all of it.
Where development is heading
Succinct proofs of source chain state, verified cheaply on the destination.
Which would remove the external attestation layer where it is deployed.
Proving cost and latency are being addressed, and adoption is early.
Interoperability standards
Common interfaces allowing applications to switch between protocols.
Which reduces lock-in and has not converged on a single approach.
Applications supporting several protocols reduce their exposure to any one failing.
A closing note
Of every category in this field, cross-chain infrastructure has the worst loss record relative to the capital it holds, and the causes have been documented after every incident.
The information required to assess a bridge before using it is public, free and takes minutes to read.
Whether anyone reads it before moving funds is the whole of the difference between the two outcomes.