[BUG] UpstashTransport buffer grows without bound during outages

Finding: In an external report against @mastra/loggers (current main at the time of the report, Node.js 24), UpstashTransport reportedly keeps failed batches in its in-memory logBuffer without any size limit and can start overlapping flush requests while the Upstash pipeline endpoint is rejecting requests, so memory and concurrent request counts can grow during a sustained outage (https://github.com/mastra-ai/mastra/issues/26489). Reported symptom: the reporter says that with a small batchSize and a consistently failing endpoint, continued writes keep growing logBuffer, and that this could exhaust application memory in a prolonged outage. The expected behavior listed by the reporter is a bounded buffer that drops the oldest records, exposes a dropped-record count, and prevents overlapping flushes (https://github.com/mastra-ai/mastra/issues/26489). Triage analysis (automated triage comment, a model-generated hypothesis rather than independent evidence): it attributes the behavior to _flush() putting failed records back with unshift when the Upstash command throws, and to _transform() calling _flush() on every write while the buffer stays at or above batchSize, with no check whether a flush is already running. It also says the interval flush adds more calls, and that overlapping successful flushes could push batches out of order. It notes that remote retention is separately capped by maxListLength (default 10,000) (https://github.com/mastra-ai/mastra/issues/26489#issuecomment-6077172013). The same comment describes a local run on main at commit 4d0056f3 with a stubbed delayed-rejecting fetch, batchSize 2, maxListLength 10, and 5,000 writes, reporting 5,000 buffered records and 2,500 fetch calls in flight at once. This is a reported result from the triage comment; KFA did not reproduce it, and it is not independent confirmation. Status: the issue is closed, and a release-notification comment says it was resolved in PR #26605 and will be included in the next release (https://github.com/mastra-ai/mastra/issues/26489#issuecomment-6100877813). The PR itself was not supplied, so the nature of the fix is unverified, and no released version is identified. Closure and a bot notice alone do not establish a shipped fix. Practical implication, conditional on the reported facts: users of UpstashTransport on main-era code who experience prolonged Upstash outages should check whether their installed @mastra/loggers release includes PR #26605 before relying on bounded memory behavior. No workaround is documented in the sources. This record is an external report with no KFA reproduction.