During syncing, we might transiently have enough malicious blaming
blocks to consider a correct header incorrect. For this reason, we
first wait for enough header to accumulate before checking blaming.
TODO(thegaram): review value and expose this as a parameter
The threshold controlling whether a node is in catch-up mode.
A node is in catch-up mode if its local best epoch number is
CATCH_UP_EPOCH_LAG_THRESHOLD behind the median of the epoch
numbers of peers.
During log filtering, we stream a set of items (blooms, receipts, txs)
to match against. To make the process faster, we need to make sure that
there’s always plenty of items in flight. This way, we can reduce idle
time when we’re waiting to receive an item.
Maximum number of in-flight epoch requests at any given time.
Similar to MAX_HEADERS_IN_FLIGHT. However, it is hard to match
hash responses to epoch requests, so we count the requests instead.