As I understand it, you're using a distributed oracle approach for moving the coins. This is the very same approach taken by the now defunct Coinsigner, Metalair, NXT Multigateway, and several other projects who proposed to use multi-signature to move coins.
I'm not familar with Metalair but I do see an organization called MetaLair.org whose business is outside the scope of cryptoasset exchange.
Coinsigner claims to be creating a dispute resolution or escrow service using 2 of 3 multisig where Coinsigner has one of the three keys. I don't see how it has any relevance to an automated cryptoasset exchange like B&C Exchange.
Multigateway appears to essentially be synonymous with SuperNET, which has some similarities to B&C Exchange but also some important differences. They are using three static multisig signers in SuperNET whereas B&C Exchange can dynamically size the number and identity of signers without any centralized control or software updates. Furthermore, SuperNET makes use of proxy assets (such mgwBTC) that introduce some difficulties. If you want to trade Bitcoin for Litecoin, for example, you must deposit Bitcoin which then entitles you to an equivalent quantity of mgwBTC which is traded for mgwLTC which entitle you to receive LTC on deposit. The supply of mgwBTC is fixed, so the number of mgwBTC will not equal the number of BTC on deposit. Excess mgwBTC that does not represent deposited BTC could be sold fraudulently, for instance. B&C Exchange does not use proxy assets and is not exposed to any of the problems and extra complications associated with it.
tl;dr; Coinsigner, MetaLair or NXT Multigateway have modest to zero relevance to B&C Exchange. Nothing that has been done by them indicates there is a problem with the design of B&C Exchange.The biggest problem with this design is it doesn't actually solve the trust problem. Yes, you now have considerably more points that can fail and in theory it should be harder to do so, but all you're basically doing is moving the problem around instead of actually solving it.
Peer to peer financial solutions establish trust by distributing responsibility among many parties and requiring transparency, consensus and verification of transactions by many parties. B&C Exchange employs these proven approaches to establishing trust.
For example: how are the signers selected? If the signers are randomly chosen then this is really no different to entrusting your money to a stranger. I mean if anyone can participate and if the system were anonymous or pseudo-anonymous then you don't know how many signers are controlled by one party. You say that reputation is the solution but it is not. Silk Road and other anonymous market places also had reputation systems to curve scamming and it was never 100% effective. Why? Because a signer can still gain reputation and use their influence to strike at an opportune time. They might create relationships with other signers and then plan an exit when their services are being used to protect substantial assets or maybe just create hundreds of signers to gain a majority control (actually, isn't this similar to Stellar?) - so this is a hard trust model to defend.
Signers are not chosen randomly. They are chosen by BlockShare holders in a dynamic fashion. Each time a BlockShare holder finds a block, they can upvote or downvote 3 reputed signers. The protocol only selects top reputed signers (those with the most points).
Reputation is a basis for most economic activity so there is no shortage of precedent here. It usually works, but not always. The system is designed to gracefully handle multiple simultaneous failures of reputed signers without loss of funds or disruption of service. In the case of an 8 of 15 multisig deposit address, the network can handle up to 7 rogue or malfunctioning signers without loss of funds. That is a very broad margin of safety for errors and malicious intent.
How do you gain reputation? When the system first launches there will be no signers with a reputation and all initial users would be gambling - i.e. the trust model would effectively be no better than a centralized exchange. You say there are enough signers where collusion and failure aren't a problem but this isn't a mathematically provable property.
Top reputed signers are likely to be established and well known pseudonymous members of the community. This means a single signer is unlikely to gain control of multiple top reputed signer identities.
Consider Mercury Exchange, for instance. Mercury uses hash-locked smart contracts to solve the trust problem. It's an approach that will be entirely trustless when transaction malleability is phased out and best of all: it won't require a third-party to mediate. But your solution on the other hand relies on the assumption that enough oracles won't be corrupted for the multi-sig to hold which isn't a provable assumption; At least with Mercury you can reason exactly about the outcomes.
It has already been pointed out that transaction malleability is a problem for Mercury's approach. It is also important that no one has a financial stake in promoting Mercury and that NuBit shareholders will provide deep liquidity on B&C Exchange but don't have the same incentives to support Mercury. I hope to have time to provide a detailed comparison of Mercury and B&C Exchange. Briefly, Mercury is an important and noble project but it doesn't have the same prospects for widespread adoption and commercial success as B&C Exchange.
There's also something you said which is quite concerning: if enough signers go offline then the money can't be moved. That's really bad. So now you have a potentially insecure and unreliable financial system which you can't prove is secure or reliable since its based on elements entirely outside your control.
Reliability of a reputed signer is perhaps their most important attribute and shareholders will upvote and downvote signers accordingly. Signers who convincingly demonstrate reliability will be upvoted (and receive income) and signers who are less than perfect will be eliminated as top reputed signers quickly. As explained earlier in this post, there is a high degree of fault tolerance (up to 7 of 15 signers can go offline without preventing funds from being unavailable in the case of an 8 of 15 deposit address).
I will address your suggested improvements later.
