Layers and Semicolons.

Jeff, yes, semicolons separate layers of a citation. As in everyday writing, they separate

• one standalone statement from another to which it is closely related; or
• items in a series that have internal commas.

When we are citing images found online, we have two approaches (see EE4 3.16):

1. When we have full data for the original:

• Layer 1: cite the imaged original, fully, using information within the images themselves—not information supplied by the provider or found within the provider's information frame around the image and not information from the provider's catalog, but only what we can verify from the image itself;
• Layer 2: cite the online provider: Creator of Database (if appropriate), “Database Title,” Creator of Website (if appropriate), Website Title (URL=Place of publication : date) > path, if appropriate > specific item/image.

2. When (a) we have incomplete data for the original or (b) otherwise need to feature the database/website:

• Layer 1: cite the online provider: Creator of Database (if appropriate), “Database Title,” Creator of Website (if appropriate), Website Title (URL=Place of publication : date) > specific item, which includes as much ID about the original document as we can discern from the image itself (not the provider's information frame around the image); 
• Layer 2: cite whatever details the website tells us about the original and its access.

This is further explained/demonstrated at 4.15 Digital Collections, under

• Online Images: Emphasis on Document
• Online Images or Transcripts: Emphasis on Database

Jeff, you write:

I think EE in the past has referred to finding that info by say backtracking through the FamilySearch catalog or sometimes the introductory images on FHL microfilm (I forget the exact term, but where it shows Item 1, Item 2,

Always, if we're using microfilm, we should backtrack to the start of the roll (or the start of that record set if there are multiple sets on the same roll) to identify what we are using. We do that when using books also; we flip back to the title page to identify what we are using. The principle holds true for using online images of microfilm.

In prior editions, EE 2.27 FHL Film of Unpublished Records stated:

Records of local, state, and federal agencies, when filmed by GSU and consulted at FHL, are cited the same way we would cite those records if we used them in their original depositories—after which we add the
FHL film number. To create our citation, we should copy precisely the label from each filmed book or file. If a register or file is not labeled, we should look for the target that the GSU imagers placed at the start
of the material.

If the target is missing or seems to be inaccurate, consult the catalog entry and create a generic label (record type and time period) using the descriptive note for that film. Be aware, however, that the FHL
catalog description frequently uses a generic label to describe the contents of an entire roll. The actual title of a specific register or file may not appear in the cataloging entry.

This approach is presented only as last-ditch effort to identify unpublished records on FHL film. It also carries a warning about the use of FHL cataloging descriptions in a way that would imply it was the actual title of the material. Earlier editions of EE carry other passages that also cautione against substituting catalog descriptions for actual labels.

The key point is that anytime we use images of anything, provided by a third party, we don't just copy the provider's identification and assume it is correct. We attempt to identity what we are using through the actual images. If we are forced to use the provider's descriptive data, we do not attach it to the details about the original as though we have indeed verified its accuracy.