The concept of "open syllable" that is relevant to pronouncing "a" as /eɪ/ is not the same as the concept of "open syllable" that is related to IPA transcriptions. You're mixing up two distinct types of analysis.
The basic idea of an "open" syllable is a syllable that ends in a vowel. Since syllabification rules for modern English pronunciation are somewhat controversial, it can sometimes be difficult to tell if a syllable is "open" in modern English pronunciation. However, as StoneyB and Janus Bahs Jacquet say in the comments, this is not the case with the words face and faced. These are unambiguously closed syllables in modern English.
But the thing that's relevant for the pronuciation of a written single vowel letter as "long" (e.g. "a" as /eɪ/) is actually whether it was in an open syllable in older forms of English (roughly speaking).
For example, we can explain the pronunciation of the word "name" something like this:
- in Middle English, the final e was pronounced (as a schwa sound), not silent
- this means it was two syllables: na-me
- This means the first syllable, "na", was "open" (it ended in a vowel, not a consonant)
- So the vowel in the first syllable was lengthened
- "a" that was lengthened in Middle English corresponds to modern English /eɪ/
This explanation is more or less true for the word "name". Some other words are spelled with silent "e" for non-historical reasons, just to mark the length of the vowel. (For example, the word wife, which was a closed syllable in Old English: wīf.) But you can pretend that they are explained the same way if you just want a rule to use to help get the modern English pronunciation.
As I said, it's important to keep in mind that syllables that were "open" in Middle English (or that we can pretend were open in Middle English) are not the same as syllables that are open in Modern English! There were a number of sound changes that occured between these, like the loss of word-final schwa sounds and the simplification of double consonant sounds. So for the purposes of determining if a letter "a" represents the "long a" sound /eɪ/ or the "short a" sound /æ/, it doesn't work to look the syllable structure in modern English. You have to see if the syllable is, not phonologically open, but what we could call "orthographically open".
To determine if the "a" in "faces" and "faced" is in an "orthographically open" syllable, you can't use IPA transcriptions. Instead, use the rule that a single consonant letter representing a single consonant sound (so, excluding tt and x, for example) should be grouped with the written syllable to its right.
Divide up the words like this: fa-ces, fa-ced. The written syllable "fa" ends in a vowel, so it is "orthographically open" and the "a" is expected to be long. The same rule applies to words like baker, navel, taken. Unsurprisingly, there are exceptions even when you apply the rule like this: e.g. gavel, water, father, panic. The rule assigning "long" pronunciations to vowel letters in orthographically open syllables is just one part of the complicated English spelling system.
As you've said, it is not regular for the inflectional suffixes -(e)s, -ed to change the length of a vowel in an English word, so that is another rule you could use. (For -(e)s, there are a few exceptions like do-does, say-says, staff-staves.) Actually, in many cases this rule gives the same result as the previous rule, because of the system of doubling consonant letters after vowel letters representing "short" stressed vowels in inflected forms. (E.g. the word knit has a "short i" in an "orthographically closed" syllable, and in the inflected forms the "short i" remains in an "orthographically closed" syllable due to the double-consonant rule: knit-ted, knit-ting.)
