The repeated destination is too distinctive to dismiss as one isolated typo. Yet the source-to-destination matches show that the bridge kept doing what the requests asked it to do. The tension lies between those two facts.
1. A contract address mistaken for a receiving address
This is the most economical explanation for the combined evidence. A token’s contract address is copied constantly: to import an asset, find its chart, verify its chain, or share it with somebody. A custom-recipient field turns that familiar string into something very different—a destination for actual funds.
Four direct Robinhood transfers are particularly useful evidence. They reached the same token address without using the Base bridge at all. That points to a broader address-handling problem that can exist in wallets and other tools too. It is compatible with people confusing “this token’s address” with “the address where I receive this token.”
The quick subsequent requests to self reinforce this reading. One holder sent 500 TIBBIR to the token, then tried 1 TIBBIR to self six minutes later. Another sent 10 and then 236 to the token before changing to self. These look like sequences an investigator would examine for trial, error and correction. The chain gives us the sequence; the interpretation is ours.
2. An interface that allowed a costly mistake
Even if the address was pasted by a user, the missing guard remains a product issue. This is a one-token bridge whose destination token is already known to the software. It could reject that exact address before requesting authorization, and the contracts could enforce the same rule independently.
Instead, the recovered path accepts a syntactically valid address, and the cross-chain warning can miss a contract that exists only on Robinhood. The selected address then passes through authorization and both bridge contracts. Several safety layers exist, but this particular destination falls between them.
That is the strongest code-supported version of the bug story: insufficient protection against an unsafe recipient. The stronger accusation—that the website itself inserted the token address—needs a different piece of evidence, and it has not appeared in the recovered builds.
3. A short-lived early frontend bug
The first day accounts for seven bridge deliveries to the token and one direct transfer. That clustering keeps the launch interface worth investigating. An early build could have behaved differently, and an abandoned script or deployment archive could still change the picture.
The likely-5-October page and the later recovered builds make a continuously broken default less compelling. So do the later addresses that switched successfully between destinations. An early bug would need to be intermittent, state-dependent, or confined to a version we have not recovered. September JavaScript, a reproducible old wallet interaction, or a sender’s retained transaction preview would be decisive next evidence.
4. Deliberate tests, alternative tools or linked accounts
Nine transfers were 10 TIBBIR or less, and three addresses repeated the token destination. Some could have been experiments. A shared tool, instructions circulating elsewhere, or several addresses controlled by one operator could also produce repetition without a universal website defect.
The much larger November movement is harder to explain as a cheap test. Its holder had four retained earlier requests to a different destination. That switch deserves attention, but there is no verified person or shared operator identity behind the addresses in this report.
What the outage theory gets wrong
A screening outage can stop a person from using the website. It does not, by itself, rewrite the recipient in a signed request. Here, the recipient was already the token contract on Base; the authorization and delivery preserved it. The expanded contract-event scan found no open/close cycles or later admin-event repair in the investigated period.
The most persuasive story is therefore a repeated unsafe-destination problem, with the website’s missing guard and warning blind spot providing a plausible route through the safeguards. Whether the original address entry was a paste mistake, an external tool or an uncaptured frontend state remains the narrow unresolved question—not the path the tokens actually took.