Rob's comment "A Strange Story" led me to write the following.
The recording of a given number extending to more than one recording session happens occasionally, ordinarily, when all takes waxed on the first session are rejected. A good example is Paul Whiteman's recording of From Monday On. The first session when this number was recorded dates from Feb 13, 1928. Three takes were waxed. Takes 1 and 2 were destroyed. Take 3 was first marked "hold conditional." Later this was crossed out with a red pencil and reclassified as "master." This take was not released until Dec 12, 1941, Vic 27688.
In the meantime, on Feb 28, 1928, Whiteman recorded takes 4-6 of the tune. Take 6 was mastered and released on Aug 13, 1928, Victor 21274. [See note 1]
Vic 21274 is then the first issue of From Monday On: it was released at the time the various takes were waxed on Feb 13 and 28, 1928. It must be noted that matrix numbers for all six takes are the same, 41689-X (X = 1-6).
What follows is a different situation. It is not uncommon for a band leader to record the same tune more than once within a relatively short time, even though the first recording was issued. This has to do with contractual obligations with different recording companies. Red Nichols was a master at this game. For example, on Aug 15, 1927, Red Nichols and His Five Pennies recorded "Feelin' No Pain" issued as Br 6326. A couple of months later, on Oct 12, 1927, Red and Miff Stompers recorded the same tune, issued on Vic 21183.
However, I don't remember having seen a record that was issued being recorded again, at a later date but within a relatively short time from the first [see note 2], for the same record label and by the same band, using the same matrix number. This, apparently, is the case of the multiple recordings of Baltimore by Syd Roy's Lyricals. They first recorded the tune in Dec 1927, perhaps earlier, and it was released on Imperial 1836. Nick tells us, The band recorded takes 3, 4 and 5 of Baltimore on January 17, 1928, of which takes 3 and 4 were issued, making four issued takes in total. So this is an example of a recording that was issued contemporaneously with the recording; nevertheless, additional takes were waxed by the same band for the same company within a few weeks of the first session, using the same matrix number (4767-X, X = 1-5).
Questions: 1. Why was this done? 2. Are other examples of recordings for which additional takes were made at a later date even though takes from the first session had been issued?
Note 1. Take 4 was classified as hold indefinitely and released on Jul 30, 1936, Victor 25368 as part of the Victor Bix Beiderbecke Memorial Album.
Note 2. For example, Whiteman first recorded San on June 9, 1924, and again, on Jan 12, 1928. This is the case of two entirely different recordings (with different matrix numbers) , not different takes of the same tune within a short time period.