See #7332.

During initial replication of the 5 initial ranges, replica reservations are sometimes declined. The error in the logs is:

I160621 20:40:59.638047 storage/reservation.go:155  could not book reservation {StoreRequestHeader:{NodeID:3 StoreID:3} FromNodeID:1 FromStoreID:1 RangeID:5 RangeSize:67108864}, not enough available reservation space (requested:67108864, reserved:201326592, maxReserved:5)

This differs from my expectation. With one node sending the reservations, I would expect that we'd see a sequence of reserve-replicate-fill. Instead, we somehow saw 5 consecutive reservations with the 5th one failing.